Misc

ACLs

swift.common.middleware.acl.acls_from_account_info(info)

Extract the account ACLs from the given account_info, and return the ACLs.

Parameters

info – a dict of the form returned by get_account_info

Returns

None (no ACL system metadata is set), or a dict of the form:: {‘admin’: […], ‘read-write’: […], ‘read-only’: […]}

Raises

ValueError – on a syntactically invalid header

swift.common.middleware.acl.clean_acl(name, value)

Returns a cleaned ACL header value, validating that it meets the formatting requirements for standard Swift ACL strings.

The ACL format is:

[item[,item...]]

Each item can be a group name to give access to or a referrer designation to grant or deny based on the HTTP Referer header.

The referrer designation format is:

.r:[-]value

The .r can also be .ref, .referer, or .referrer; though it will be shortened to just .r for decreased character count usage.

The value can be * to specify any referrer host is allowed access, a specific host name like www.example.com, or if it has a leading period . or leading *. it is a domain name specification, like .example.com or *.example.com. The leading minus sign - indicates referrer hosts that should be denied access.

Referrer access is applied in the order they are specified. For example, .r:.example.com,.r:-thief.example.com would allow all hosts ending with .example.com except for the specific host thief.example.com.

Example valid ACLs:

.r:*
.r:*,.r:-.thief.com
.r:*,.r:.example.com,.r:-thief.example.com
.r:*,.r:-.thief.com,bobs_account,sues_account:sue
bobs_account,sues_account:sue

Example invalid ACLs:

.r:
.r:-

By default, allowing read access via .r will not allow listing objects in the container – just retrieving objects from the container. To turn on listings, use the .rlistings directive.

Also, .r designations aren’t allowed in headers whose names include the word ‘write’.

ACLs that are “messy” will be cleaned up. Examples:

Original

Cleaned

bob, sue

bob,sue

bob , sue

bob,sue

bob,,,sue

bob,sue

.referrer : *

.r:*

.ref:*.example.com

.r:.example.com

.r:*, .rlistings

.r:*,.rlistings

Parameters
  • name – The name of the header being cleaned, such as X-Container-Read or X-Container-Write.

  • value – The value of the header being cleaned.

Returns

The value, cleaned of extraneous formatting.

Raises

ValueError – If the value does not meet the ACL formatting requirements; the error message will indicate why.

swift.common.middleware.acl.format_acl(version=1, **kwargs)

Compatibility wrapper to help migrate ACL syntax from version 1 to 2. Delegates to the appropriate version-specific format_acl method, defaulting to version 1 for backward compatibility.

Parameters

kwargs – keyword args appropriate for the selected ACL syntax version (see format_acl_v1() or format_acl_v2())

swift.common.middleware.acl.format_acl_v1(groups=None, referrers=None, header_name=None)

Returns a standard Swift ACL string for the given inputs.

Caller is responsible for ensuring that :referrers: parameter is only given if the ACL is being generated for X-Container-Read. (X-Container-Write and the account ACL headers don’t support referrers.)

Parameters
  • groups – a list of groups (and/or members in most auth systems) to grant access

  • referrers – a list of referrer designations (without the leading .r:)

  • header_name – (optional) header name of the ACL we’re preparing, for clean_acl; if None, returned ACL won’t be cleaned

Returns

a Swift ACL string for use in X-Container-{Read,Write}, X-Account-Access-Control, etc.

swift.common.middleware.acl.format_acl_v2(acl_dict)

Returns a version-2 Swift ACL JSON string.

HTTP headers for Version 2 ACLs have the following form:

Header-Name: {“arbitrary”:”json”,”encoded”:”string”}

JSON will be forced ASCII (containing six-char uNNNN sequences rather than UTF-8; UTF-8 is valid JSON but clients vary in their support for UTF-8 headers), and without extraneous whitespace.

Advantages over V1: forward compatibility (new keys don’t cause parsing exceptions); Unicode support; no reserved words (you can have a user named .rlistings if you want).

Parameters

acl_dict – dict of arbitrary data to put in the ACL; see specific auth systems such as tempauth for supported values

Returns

a JSON string which encodes the ACL

swift.common.middleware.acl.parse_acl(*args, **kwargs)

Compatibility wrapper to help migrate ACL syntax from version 1 to 2. Delegates to the appropriate version-specific parse_acl method, attempting to determine the version from the types of args/kwargs.

Parameters
  • args – positional args for the selected ACL syntax version

  • kwargs – keyword args for the selected ACL syntax version (see parse_acl_v1() or parse_acl_v2())

Returns

the return value of parse_acl_v1() or parse_acl_v2()

swift.common.middleware.acl.parse_acl_v1(acl_string)

Parses a standard Swift ACL string into a referrers list and groups list.

See clean_acl() for documentation of the standard Swift ACL format.

Parameters

acl_string – The standard Swift ACL string to parse.

Returns

A tuple of (referrers, groups) where referrers is a list of referrer designations (without the leading .r:) and groups is a list of groups to allow access.

swift.common.middleware.acl.parse_acl_v2(data)

Parses a version-2 Swift ACL string and returns a dict of ACL info.

Parameters

data – string containing the ACL data in JSON format

Returns

A dict (possibly empty) containing ACL info, e.g.: {“groups”: […], “referrers”: […]}

Returns

None if data is None, is not valid JSON or does not parse as a dict

Returns

empty dictionary if data is an empty string

swift.common.middleware.acl.referrer_allowed(referrer, referrer_acl)

Returns True if the referrer should be allowed based on the referrer_acl list (as returned by parse_acl()).

See clean_acl() for documentation of the standard Swift ACL format.

Parameters
  • referrer – The value of the HTTP Referer header.

  • referrer_acl – The list of referrer designations as returned by parse_acl().

Returns

True if the referrer should be allowed; False if not.

Buffered HTTP

Monkey Patch httplib.HTTPResponse to buffer reads of headers. This can improve performance when making large numbers of small HTTP requests. This module also provides helper functions to make HTTP connections using BufferedHTTPResponse.

Warning

If you use this, be sure that the libraries you are using do not access the socket directly (xmlrpclib, I’m looking at you :/), and instead make all calls through httplib.

class swift.common.bufferedhttp.BufferedHTTPConnection(host, port=None, timeout=<object object>, source_address=None)

Bases: eventlet.green.http.client.HTTPConnection

HTTPConnection class that uses BufferedHTTPResponse

connect()

Connect to the host and port specified in __init__.

getresponse()

Get the response from the server.

If the HTTPConnection is in the correct state, returns an instance of HTTPResponse or of whatever object is returned by the response_class variable.

If a request has not been sent or if a previous response has not be handled, ResponseNotReady is raised. If the HTTP response indicates that the connection should be closed, then it will be closed before the response is returned. When the connection is closed, the underlying socket is closed.

putheader(header, value)

Send a request header line to the server.

For example: h.putheader(‘Accept’, ‘text/html’)

putrequest(method, url, skip_host=0, skip_accept_encoding=0)

Send a request to the server.

Parameters
  • method – specifies an HTTP request method, e.g. ‘GET’.

  • url – specifies the object being requested, e.g. ‘/index.html’.

  • skip_host – if True does not add automatically a ‘Host:’ header

  • skip_accept_encoding – if True does not add automatically an ‘Accept-Encoding:’ header

response_class

alias of swift.common.bufferedhttp.BufferedHTTPResponse

class swift.common.bufferedhttp.BufferedHTTPResponse(sock, debuglevel=0, strict=0, method=None)

Bases: eventlet.green.http.client.HTTPResponse

HTTPResponse class that buffers reading of headers

close()

Flush and close the IO object.

This method has no effect if the file is already closed.

nuke_from_orbit()

Terminate the socket with extreme prejudice.

Closes the underlying socket regardless of whether or not anyone else has references to it. Use this when you are certain that nobody else you care about has a reference to this socket.

read(amt=None)

Read and return up to n bytes.

If the argument is omitted, None, or negative, reads and returns all data until EOF.

If the argument is positive, and the underlying raw stream is not ‘interactive’, multiple raw reads may be issued to satisfy the byte count (unless EOF is reached first). But for interactive raw streams (as well as sockets and pipes), at most one raw read will be issued, and a short result does not imply that EOF is imminent.

Returns an empty bytes object on EOF.

Returns None if the underlying raw stream was open in non-blocking mode and no data is available at the moment.

readline(size=1024)

Read and return a line from the stream.

If size is specified, at most size bytes will be read.

The line terminator is always b’n’ for binary files; for text files, the newlines argument to open can be used to select the line terminator(s) recognized.

swift.common.bufferedhttp.http_connect(ipaddr, port, device, partition, method, path, headers=None, query_string=None, ssl=False)

Helper function to create an HTTPConnection object. If ssl is set True, HTTPSConnection will be used. However, if ssl=False, BufferedHTTPConnection will be used, which is buffered for backend Swift services.

Parameters
  • ipaddr – IPv4 address to connect to

  • port – port to connect to

  • device – device of the node to query

  • partition – partition on the device

  • method – HTTP method to request (‘GET’, ‘PUT’, ‘POST’, etc.)

  • path – request path

  • headers – dictionary of headers

  • query_string – request query string

  • ssl – set True if SSL should be used (default: False)

Returns

HTTPConnection object

swift.common.bufferedhttp.http_connect_raw(ipaddr, port, method, path, headers=None, query_string=None, ssl=False)

Helper function to create an HTTPConnection object. If ssl is set True, HTTPSConnection will be used. However, if ssl=False, BufferedHTTPConnection will be used, which is buffered for backend Swift services.

Parameters
  • ipaddr – IPv4 address to connect to

  • port – port to connect to

  • method – HTTP method to request (‘GET’, ‘PUT’, ‘POST’, etc.)

  • path – request path

  • headers – dictionary of headers

  • query_string – request query string

  • ssl – set True if SSL should be used (default: False)

Returns

HTTPConnection object

Constraints

swift.common.constraints.check_account_format(req, name, *, target_type='Account')

Validate that the header contains valid account or container name.

Parameters
  • req – HTTP request object

  • name – header value to validate

  • target_type – which header is being validated (Account or Container)

Returns

A properly encoded account name or container name

Raises

HTTPPreconditionFailed – if account header is not well formatted.

swift.common.constraints.check_container_format(req, name, *, target_type='Container')

Validate that the header contains valid account or container name.

Parameters
  • req – HTTP request object

  • name – header value to validate

  • target_type – which header is being validated (Account or Container)

Returns

A properly encoded account name or container name

Raises

HTTPPreconditionFailed – if account header is not well formatted.

swift.common.constraints.check_delete_headers(request)

Check that ‘x-delete-after’ and ‘x-delete-at’ headers have valid values. Values should be positive integers and correspond to a time greater than the request timestamp.

If the ‘x-delete-after’ header is found then its value is used to compute an ‘x-delete-at’ value which takes precedence over any existing ‘x-delete-at’ header.

Parameters

request – the swob request object

Raises

HTTPBadRequest in case of invalid values

Returns

the swob request object

swift.common.constraints.check_dir(root, drive)

Verify that the path to the device is a directory and is a lesser constraint that is enforced when a full mount_check isn’t possible with, for instance, a VM using loopback or partitions.

Parameters
  • root – base path where the dir is

  • drive – drive name to be checked

Returns

full path to the device

Raises

ValueError – if drive fails to validate

swift.common.constraints.check_drive(root, drive, mount_check)

Validate the path given by root and drive is a valid existing directory.

Parameters
  • root – base path where the devices are mounted

  • drive – drive name to be checked

  • mount_check – additionally require path is mounted

Returns

full path to the device

Raises

ValueError – if drive fails to validate

swift.common.constraints.check_float(string)

Helper function for checking if a string can be converted to a float.

Parameters

string – string to be verified as a float

Returns

True if the string can be converted to a float, False otherwise

swift.common.constraints.check_metadata(req, target_type)

Check metadata sent in the request headers. This should only check that the metadata in the request given is valid. Checks against account/container overall metadata should be forwarded on to its respective server to be checked.

Parameters
  • req – request object

  • target_type – str: one of: object, container, or account: indicates which type the target storage for the metadata is

Returns

HTTPBadRequest with bad metadata otherwise None

swift.common.constraints.check_mount(root, drive)

Verify that the path to the device is a mount point and mounted. This allows us to fast fail on drives that have been unmounted because of issues, and also prevents us for accidentally filling up the root partition.

Parameters
  • root – base path where the devices are mounted

  • drive – drive name to be checked

Returns

full path to the device

Raises

ValueError – if drive fails to validate

swift.common.constraints.check_name_format(req, name, target_type)

Validate that the header contains valid account or container name.

Parameters
  • req – HTTP request object

  • name – header value to validate

  • target_type – which header is being validated (Account or Container)

Returns

A properly encoded account name or container name

Raises

HTTPPreconditionFailed – if account header is not well formatted.

swift.common.constraints.check_object_creation(req, object_name)

Check to ensure that everything is alright about an object to be created.

Parameters
  • req – HTTP request object

  • object_name – name of object to be created

Returns

HTTPRequestEntityTooLarge – the object is too large

Returns

HTTPLengthRequired – missing content-length header and not a chunked request

Returns

HTTPBadRequest – missing or bad content-type header, or bad metadata

Returns

HTTPNotImplemented – unsupported transfer-encoding header value

swift.common.constraints.check_utf8(string, internal=False)

Validate if a string is valid UTF-8 str or unicode and that it does not contain any reserved characters.

Parameters
  • string – string to be validated

  • internal – boolean, allows reserved characters if True

Returns

True if the string is valid utf-8 str or unicode and contains no null characters, False otherwise

swift.common.constraints.reload_constraints()

Parse SWIFT_CONF_FILE and reset module level global constraint attrs, populating OVERRIDE_CONSTRAINTS AND EFFECTIVE_CONSTRAINTS along the way.

swift.common.constraints.valid_api_version(version)

Checks if the requested version is valid.

Currently Swift only supports “v1” and “v1.0”.

swift.common.constraints.valid_timestamp(request)

Helper function to extract a timestamp from requests that require one.

Parameters

request – the swob request object

Returns

a valid Timestamp instance

Raises

HTTPBadRequest – on missing or invalid X-Timestamp

Container Sync Realms

class swift.common.container_sync_realms.ContainerSyncRealms(conf_path, logger)

Bases: object

Loads and parses the container-sync-realms.conf, occasionally checking the file’s mtime to see if it needs to be reloaded.

clusters(realm)

Returns a list of clusters for the realm.

endpoint(realm, cluster)

Returns the endpoint for the cluster in the realm.

get_sig(request_method, path, x_timestamp, nonce, realm_key, user_key)

Returns the hexdigest string of the HMAC-SHA1 (RFC 2104) for the information given.

Parameters
  • request_method – HTTP method of the request.

  • path – The path to the resource (url-encoded).

  • x_timestamp – The X-Timestamp header value for the request.

  • nonce – A unique value for the request.

  • realm_key – Shared secret at the cluster operator level.

  • user_key – Shared secret at the user’s container level.

Returns

hexdigest str of the HMAC-SHA1 for the request.

key(realm)

Returns the key for the realm.

key2(realm)

Returns the key2 for the realm.

realms()

Returns a list of realms.

reload()

Forces a reload of the conf file.

Digest

swift.common.digest.extract_digest_and_algorithm(value)

Returns a tuple of (digest_algorithm, hex_encoded_digest) from a client-provided string of the form:

<hex-encoded digest>

or:

<algorithm>:<base64-encoded digest>

Note that hex-encoded strings must use one of sha1, sha256, or sha512.

Raises

ValueError on parse failures

swift.common.digest.get_allowed_digests(conf_digests, logger=None)

Pulls out ‘allowed_digests’ from the supplied conf. Then compares them with the list of supported and deprecated digests and returns whatever remain.

When something is unsupported or deprecated it’ll log a warning.

Parameters
  • conf_digests – iterable of allowed digests. If empty, defaults to DEFAULT_ALLOWED_DIGESTS.

  • logger – optional logger; if provided, use it issue deprecation warnings

Returns

A set of allowed digests that are supported and a set of deprecated digests.

Raises

ValueError, if there are no digests left to return.

swift.common.digest.get_hmac(request_method, path, expires, key, digest='sha1', ip_range=None)

Returns the hexdigest string of the HMAC (see RFC 2104) for the request.

Parameters
  • request_method – Request method to allow.

  • path – The path to the resource to allow access to.

  • expires – Unix timestamp as an int for when the URL expires.

  • key – HMAC shared secret.

  • digest – constructor or the string name for the digest to use in calculating the HMAC Defaults to SHA1

  • ip_range – The ip range from which the resource is allowed to be accessed. We need to put the ip_range as the first argument to hmac to avoid manipulation of the path due to newlines being valid in paths e.g. /v1/a/c/on127.0.0.1

Returns

hexdigest str of the HMAC for the request using the specified digest algorithm.

Direct Client

Internal client library for making calls directly to the servers rather than through the proxy.

exception swift.common.direct_client.DirectClientException(stype, method, node, part, path, resp, host=None)

Bases: swift.common.exceptions.ClientException

exception swift.common.direct_client.DirectClientReconException(method, node, path, resp)

Bases: swift.common.exceptions.ClientException

swift.common.direct_client.direct_delete_account(node, part, account, conn_timeout=5, response_timeout=15, headers=None)
swift.common.direct_client.direct_delete_container(node, part, account, container, conn_timeout=5, response_timeout=15, headers=None)

Delete container directly from the container server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • headers – dict to be passed into HTTPConnection headers

Raises

ClientException – HTTP DELETE request failed

swift.common.direct_client.direct_delete_container_object(node, part, account, container, obj, conn_timeout=5, response_timeout=15, headers=None)
swift.common.direct_client.direct_delete_object(node, part, account, container, obj, conn_timeout=5, response_timeout=15, headers=None)

Delete object directly from the object server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • obj – object name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

Raises

ClientException – HTTP DELETE request failed

swift.common.direct_client.direct_get_account(node, part, account, marker=None, limit=None, prefix=None, delimiter=None, conn_timeout=5, response_timeout=15, end_marker=None, reverse=None, headers=None)

Get listings directly from the account server.

Parameters
  • node – node dictionary from the ring

  • part – partition the account is on

  • account – account name

  • marker – marker query

  • limit – query limit

  • prefix – prefix query

  • delimiter – delimiter for the query

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • end_marker – end_marker query

  • reverse – reverse the returned listing

Returns

a tuple of (response headers, a list of containers) The response headers will HeaderKeyDict.

swift.common.direct_client.direct_get_container(node, part, account, container, marker=None, limit=None, prefix=None, delimiter=None, conn_timeout=5, response_timeout=15, end_marker=None, reverse=None, headers=None)

Get container listings directly from the container server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • marker – marker query

  • limit – query limit

  • prefix – prefix query

  • delimiter – delimiter for the query

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • end_marker – end_marker query

  • reverse – reverse the returned listing

  • headers – headers to be included in the request

Returns

a tuple of (response headers, a list of objects) The response headers will be a HeaderKeyDict.

swift.common.direct_client.direct_get_object(node, part, account, container, obj, conn_timeout=5, response_timeout=15, resp_chunk_size=None, headers=None)

Get object directly from the object server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • obj – object name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • resp_chunk_size – if defined, chunk size of data to read.

  • headers – dict to be passed into HTTPConnection headers

Returns

a tuple of (response headers, the object’s contents) The response headers will be a HeaderKeyDict.

Raises

ClientException – HTTP GET request failed

swift.common.direct_client.direct_get_recon(node, recon_command, conn_timeout=5, response_timeout=15, headers=None)

Get recon json directly from the storage server.

Parameters
  • node – node dictionary from the ring

  • recon_command – recon string (post /recon/)

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • headers – dict to be passed into HTTPConnection headers

Returns

deserialized json response

Raises

DirectClientReconException – HTTP GET request failed

swift.common.direct_client.direct_get_suffix_hashes(node, part, suffixes, conn_timeout=5, response_timeout=15, headers=None)

Get suffix hashes directly from the object server.

Note that unlike other direct_client functions, this one defaults to using the replication network to make requests.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • headers – dict to be passed into HTTPConnection headers

Returns

dict of suffix hashes

Raises

ClientException – HTTP REPLICATE request failed

swift.common.direct_client.direct_head_container(node, part, account, container, conn_timeout=5, response_timeout=15, headers=None)

Request container information directly from the container server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

Returns

a dict containing the response’s headers in a HeaderKeyDict

Raises

ClientException – HTTP HEAD request failed

swift.common.direct_client.direct_head_object(node, part, account, container, obj, conn_timeout=5, response_timeout=15, headers=None)

Request object information directly from the object server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • obj – object name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • headers – dict to be passed into HTTPConnection headers

Returns

a dict containing the response’s headers in a HeaderKeyDict

Raises

ClientException – HTTP HEAD request failed

swift.common.direct_client.direct_post_object(node, part, account, container, name, headers, conn_timeout=5, response_timeout=15)

Direct update to object metadata on object server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • name – object name

  • headers – headers to store as metadata

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

Raises

ClientException – HTTP POST request failed

swift.common.direct_client.direct_put_container(node, part, account, container, conn_timeout=5, response_timeout=15, headers=None, contents=None, content_length=None, chunk_size=65535)

Make a PUT request to a container server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • headers – additional headers to include in the request

  • contents – an iterable or string to send in request body (optional)

  • content_length – value to send as content-length header (optional)

  • chunk_size – chunk size of data to send (optional)

Raises

ClientException – HTTP PUT request failed

swift.common.direct_client.direct_put_container_object(node, part, account, container, obj, conn_timeout=5, response_timeout=15, headers=None)
swift.common.direct_client.direct_put_object(node, part, account, container, name, contents, content_length=None, etag=None, content_type=None, headers=None, conn_timeout=5, response_timeout=15, chunk_size=65535)

Put object directly from the object server.

Parameters
  • node – node dictionary from the ring

  • part – partition the container is on

  • account – account name

  • container – container name

  • name – object name

  • contents – an iterable or string to read object data from

  • content_length – value to send as content-length header

  • etag – etag of contents

  • content_type – value to send as content-type header

  • headers – additional headers to include in the request

  • conn_timeout – timeout in seconds for establishing the connection

  • response_timeout – timeout in seconds for getting the response

  • chunk_size – if defined, chunk size of data to send.

Returns

etag from the server response

Raises

ClientException – HTTP PUT request failed

swift.common.direct_client.gen_headers(hdrs_in=None, add_ts=True)

Get the headers ready for a request. All requests should have a User-Agent string, but if one is passed in don’t over-write it. Not all requests will need an X-Timestamp, but if one is passed in do not over-write it.

Parameters
  • headers – dict or None, base for HTTP headers

  • add_ts – boolean, should be True for any “unsafe” HTTP request

Returns

HeaderKeyDict based on headers and ready for the request

swift.common.direct_client.retry(func, *args, **kwargs)

Helper function to retry a given function a number of times.

Parameters
  • func – callable to be called

  • retries – number of retries

  • error_log – logger for errors

  • args – arguments to send to func

  • kwargs – keyward arguments to send to func (if retries or error_log are sent, they will be deleted from kwargs before sending on to func)

Returns

result of func

Raises

ClientException – all retries failed

Exceptions

exception swift.common.exceptions.APIVersionError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ChunkReadError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ChunkReadTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

exception swift.common.exceptions.ChunkWriteTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

exception swift.common.exceptions.ClientException(msg, http_scheme='', http_host='', http_port='', http_path='', http_query='', http_status=None, http_reason='', http_device='', http_response_content='', http_headers=None)

Bases: Exception

exception swift.common.exceptions.ConnectionTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

exception swift.common.exceptions.DatabaseAuditorException

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.DeviceUnavailable

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.DiskFileBadMetadataChecksum

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileCollision

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileDeleted(metadata=None)

Bases: swift.common.exceptions.DiskFileNotExist

exception swift.common.exceptions.DiskFileDeviceUnavailable

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.DiskFileExpired(metadata=None)

Bases: swift.common.exceptions.DiskFileDeleted

exception swift.common.exceptions.DiskFileNoSpace

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileNotExist

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileNotOpen

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileQuarantined

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DiskFileXattrNotSupported

Bases: swift.common.exceptions.DiskFileError

exception swift.common.exceptions.DriveNotMounted

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.DuplicateDeviceError

Bases: swift.common.exceptions.RingBuilderError

exception swift.common.exceptions.EmptyRingError

Bases: swift.common.exceptions.RingBuilderError

exception swift.common.exceptions.EncryptionException

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.FileNotFoundError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.FooterNotSupported

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.InsufficientStorage

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.InvalidAccountInfo

Bases: swift.common.exceptions.DatabaseAuditorException

exception swift.common.exceptions.InvalidPidFileException

Bases: Exception

exception swift.common.exceptions.InvalidTimestamp

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.LinkIterError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ListingIterError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ListingIterNotAuthorized(aresp)

Bases: swift.common.exceptions.ListingIterError

exception swift.common.exceptions.ListingIterNotFound

Bases: swift.common.exceptions.ListingIterError

exception swift.common.exceptions.LockTimeout(seconds=None, msg=None)

Bases: swift.common.exceptions.MessageTimeout

exception swift.common.exceptions.MessageTimeout(seconds=None, msg=None)

Bases: eventlet.timeout.Timeout

exception swift.common.exceptions.MimeInvalid

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.MultiphasePUTNotSupported

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.PartitionLockTimeout(seconds=None, msg=None)

Bases: swift.common.exceptions.LockTimeout

exception swift.common.exceptions.PathNotDir

Bases: OSError

exception swift.common.exceptions.PermissionError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.PutterConnectError(status=None)

Bases: Exception

exception swift.common.exceptions.QuarantineRequest

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.RangeAlreadyComplete

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ReplicationException

Bases: Exception

exception swift.common.exceptions.ReplicationLockTimeout(seconds=None, msg=None)

Bases: swift.common.exceptions.LockTimeout

exception swift.common.exceptions.ResponseTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

exception swift.common.exceptions.RingBuilderError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.RingLoadError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.RingValidationError

Bases: swift.common.exceptions.RingBuilderError

exception swift.common.exceptions.SegmentError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.ShortReadError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.SuffixSyncError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.SwiftException

Bases: Exception

exception swift.common.exceptions.UnPicklingError

Bases: swift.common.exceptions.SwiftException

exception swift.common.exceptions.UnknownSecretIdError

Bases: swift.common.exceptions.EncryptionException

Internal Client

class swift.common.internal_client.CompressingFileReader(file_obj, compresslevel=9, chunk_size=4096)

Bases: object

Wrapper for file object to compress object while reading.

Can be used to wrap file objects passed to InternalClient.upload_object().

Used in testing of InternalClient.

Parameters
  • file_obj – File object to wrap.

  • compresslevel – Compression level, defaults to 9.

  • chunk_size – Size of chunks read when iterating using object, defaults to 4096.

next()
read(*a, **kw)

Reads a chunk from the file object.

Params are passed directly to the underlying file object’s read().

Returns

Compressed chunk from file object.

seek(offset, whence=0)
set_initial_state()

Sets the object to the state needed for the first read.

class swift.common.internal_client.InternalClient(conf_path, user_agent, request_tries, allow_modify_pipeline=False, use_replication_network=False, global_conf=None, app=None)

Bases: object

An internal client that uses a swift proxy app to make requests to Swift.

This client will exponentially slow down for retries.

Parameters
  • conf_path – Full path to proxy config.

  • user_agent – User agent to be sent to requests to Swift.

  • request_tries – Number of tries before InternalClient.make_request() gives up.

  • global_conf – a dict of options to update the loaded proxy config. Options in global_conf will override those in conf_path except where the conf_path option is preceded by set.

  • app – Optionally provide a WSGI app for the internal client to use.

container_exists(account, container)

Checks to see if a container exists.

Parameters
  • account – The container’s account.

  • container – Container to check.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

Returns

True if container exists, false otherwise.

create_account(account)

Creates an account.

Parameters

account – Account to create.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

create_container(account, container, headers=None, acceptable_statuses=(2,))

Creates container.

Parameters
  • account – The container’s account.

  • container – Container to create.

  • headers – Defaults to empty dict.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

delete_account(account, acceptable_statuses=(2, 404))

Deletes an account.

Parameters
  • account – Account to delete.

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

delete_container(account, container, headers=None, acceptable_statuses=(2, 404))

Deletes a container.

Parameters
  • account – The container’s account.

  • container – Container to delete.

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

delete_object(account, container, obj, acceptable_statuses=(2, 404), headers=None)

Deletes an object.

Parameters
  • account – The object’s account.

  • container – The object’s container.

  • obj – The object.

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

  • headers – extra headers to send with request

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

get_account_info(account, acceptable_statuses=(2, 404))

Returns (container_count, object_count) for an account.

Parameters
  • account – Account on which to get the information.

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

get_account_metadata(account, metadata_prefix='', acceptable_statuses=(2,), params=None)

Gets account metadata.

Parameters
  • account – Account on which to get the metadata.

  • metadata_prefix – Used to filter values from the headers returned. Will strip that prefix from the keys in the dict returned. Defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Returns

Returns dict of account metadata. Keys will be lowercase.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

get_container_metadata(account, container, metadata_prefix='', acceptable_statuses=(2,), params=None)

Gets container metadata.

Parameters
  • account – The container’s account.

  • container – Container to get metadata on.

  • metadata_prefix – Used to filter values from the headers returned. Will strip that prefix from the keys in the dict returned. Defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Returns

Returns dict of container metadata. Keys will be lowercase.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

get_object(account, container, obj, headers=None, acceptable_statuses=(2,), params=None)

Gets an object.

Parameters
  • account – The object’s account.

  • container – The object’s container.

  • obj – The object name.

  • headers – Headers to send with request, defaults to empty dict.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

  • params – A dict of params to be set in request query string, defaults to None.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

Returns

A 3-tuple (status, headers, iterator of object body)

get_object_metadata(account, container, obj, metadata_prefix='', acceptable_statuses=(2,), headers=None, params=None)

Gets object metadata.

Parameters
  • account – The object’s account.

  • container – The object’s container.

  • obj – The object.

  • metadata_prefix – Used to filter values from the headers returned. Will strip that prefix from the keys in the dict returned. Defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

  • headers – extra headers to send with request

Returns

Dict of object metadata.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

handle_request(*args, **kwargs)
iter_containers(account, marker='', end_marker='', prefix='', acceptable_statuses=(2, 404))

Returns an iterator of containers dicts from an account.

Parameters
  • account – Account on which to do the container listing.

  • marker – Prefix of first desired item, defaults to ‘’.

  • end_marker – Last item returned will be ‘less’ than this, defaults to ‘’.

  • prefix – Prefix of containers

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

iter_object_lines(account, container, obj, headers=None, acceptable_statuses=(2,))

Returns an iterator of object lines from an uncompressed or compressed text object.

Uncompress object as it is read if the object’s name ends with ‘.gz’.

Parameters
  • account – The object’s account.

  • container – The object’s container.

  • obj – The object.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

iter_objects(account, container, marker='', end_marker='', prefix='', acceptable_statuses=(2, 404))

Returns an iterator of object dicts from a container.

Parameters
  • account – The container’s account.

  • container – Container to iterate objects on.

  • marker – Prefix of first desired item, defaults to ‘’.

  • end_marker – Last item returned will be ‘less’ than this, defaults to ‘’.

  • prefix – Prefix of objects

  • acceptable_statuses – List of status for valid responses, defaults to (2, HTTP_NOT_FOUND).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

make_path(account, container=None, obj=None)

Returns a swift path for a request quoting and utf-8 encoding the path parts as need be.

Parameters
  • account – swift account

  • container – container, defaults to None

  • obj – object, defaults to None

Raises

ValueError – Is raised if obj is specified and container is not.

make_request(method, path, headers, acceptable_statuses, body_file=None, params=None)

Makes a request to Swift with retries.

Parameters
  • method – HTTP method of request.

  • path – Path of request.

  • headers – Headers to be sent with request.

  • acceptable_statuses – List of acceptable statuses for request.

  • body_file – Body file to be passed along with request, defaults to None.

  • params – A dict of params to be set in request query string, defaults to None.

Returns

Response object on success.

Raises
  • UnexpectedResponse – Exception raised when make_request() fails to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

set_account_metadata(account, metadata, metadata_prefix='', acceptable_statuses=(2,))

Sets account metadata. A call to this will add to the account metadata and not overwrite all of it with values in the metadata dict. To clear an account metadata value, pass an empty string as the value for the key in the metadata dict.

Parameters
  • account – Account on which to get the metadata.

  • metadata – Dict of metadata to set.

  • metadata_prefix – Prefix used to set metadata values in headers of requests, used to prefix keys in metadata when setting metadata, defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

set_container_metadata(account, container, metadata, metadata_prefix='', acceptable_statuses=(2,))

Sets container metadata. A call to this will add to the container metadata and not overwrite all of it with values in the metadata dict. To clear a container metadata value, pass an empty string as the value for the key in the metadata dict.

Parameters
  • account – The container’s account.

  • container – Container to set metadata on.

  • metadata – Dict of metadata to set.

  • metadata_prefix – Prefix used to set metadata values in headers of requests, used to prefix keys in metadata when setting metadata, defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

set_object_metadata(account, container, obj, metadata, metadata_prefix='', acceptable_statuses=(2,))

Sets an object’s metadata. The object’s metadata will be overwritten by the values in the metadata dict.

Parameters
  • account – The object’s account.

  • container – The object’s container.

  • obj – The object.

  • metadata – Dict of metadata to set.

  • metadata_prefix – Prefix used to set metadata values in headers of requests, used to prefix keys in metadata when setting metadata, defaults to ‘’.

  • acceptable_statuses – List of status for valid responses, defaults to (2,).

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

upload_object(fobj, account, container, obj, headers=None, acceptable_statuses=(2,), params=None)
Parameters
  • fobj – File object to read object’s content from.

  • account – The object’s account.

  • container – The object’s container.

  • obj – The object.

  • headers – Headers to send with request, defaults to empty dict.

  • acceptable_statuses – List of acceptable statuses for request.

  • params – A dict of params to be set in request query string, defaults to None.

Raises
  • UnexpectedResponse – Exception raised when requests fail to get a response with an acceptable status

  • Exception – Exception is raised when code fails in an unexpected way.

class swift.common.internal_client.SimpleClient(url=None, token=None, starting_backoff=1, max_backoff=5, retries=5)

Bases: object

Simple client that is used in bin/swift-dispersion-* and container sync

base_request(method, container=None, name=None, prefix=None, headers=None, proxy=None, contents=None, full_listing=None, logger=None, additional_info=None, timeout=None, marker=None)
get_account(*args, **kwargs)
get_container(container, **kwargs)
put_container(container, **kwargs)
put_object(container, name, contents, **kwargs)
retry_request(method, **kwargs)
exception swift.common.internal_client.UnexpectedResponse(message, resp)

Bases: Exception

Exception raised on invalid responses to InternalClient.make_request().

Parameters
  • message – Exception message.

  • resp – The unexpected response.

swift.common.internal_client.delete_object(url, **kwargs)

For usage with container sync

swift.common.internal_client.get_auth(url, user, key, auth_version='1.0', **kwargs)
swift.common.internal_client.head_object(url, **kwargs)

For usage with container sync

swift.common.internal_client.put_object(url, **kwargs)

For usage with container sync

Manager

class swift.common.manager.Manager(servers, run_dir='/var/run/swift')

Bases: object

Main class for performing commands on groups of servers.

Parameters

servers – list of server names as strings

force_reload(**kwargs)

alias for reload

get_command(cmd)

Find and return the decorated method named like cmd

Parameters

cmd – the command to get, a string, if not found raises UnknownCommandError

kill(**kwargs)

stop a server (no error if not running)

kill_child_pids(**kwargs)

kill child pids, optionally servicing accepted connections

classmethod list_commands()

Get all publicly accessible commands

Returns

a list of string tuples (cmd, help), the method names who are decorated as commands

no_daemon(**kwargs)

start a server interactively

no_wait(**kwargs)

spawn server and return immediately

once(**kwargs)

start server and run one pass on supporting daemons

reload(**kwargs)

graceful shutdown then restart on supporting servers

reload_seamless(**kwargs)

seamlessly re-exec, then shutdown of old listen sockets on supporting servers

restart(**kwargs)

stops then restarts server

run_command(cmd, **kwargs)

Find the named command and run it

Parameters

cmd – the command name to run

shutdown(**kwargs)

allow current requests to finish on supporting servers

start(**kwargs)

starts a server

status(**kwargs)

display status of tracked pids for server

stop(**kwargs)

stops a server

class swift.common.manager.Server(server, run_dir='/var/run/swift')

Bases: object

Manage operations on a server or group of servers of similar type

Parameters

server – name of server

conf_files(**kwargs)

Get conf files for this server

Parameters

number – if supplied will only lookup the nth server

Returns

list of conf files

get_conf_file_name(pid_file)

Translate pid_file to a corresponding conf_file

Parameters

pid_file – a pid_file for this server, a string

Returns

the conf_file for this pid_file

get_pid_file_name(conf_file)

Translate conf_file to a corresponding pid_file

Parameters

conf_file – an conf_file for this server, a string

Returns

the pid_file for this conf_file

get_running_pids(**kwargs)

Get running pids

Returns

a dict mapping pids (ints) to pid_files (paths)

interact(**kwargs)

wait on spawned procs to terminate

iter_pid_files(**kwargs)

Generator, yields (pid_file, pids)

kill_child_pids(**kwargs)

Kill child pids, leaving server overseer to respawn them

Parameters
  • graceful – if True, attempt SIGHUP on supporting servers

  • seamless – if True, attempt SIGUSR1 on supporting servers

Returns

a dict mapping pids (ints) to pid_files (paths)

kill_running_pids(**kwargs)

Kill running pids

Parameters
  • graceful – if True, attempt SIGHUP on supporting servers

  • seamless – if True, attempt SIGUSR1 on supporting servers

Returns

a dict mapping pids (ints) to pid_files (paths)

launch(**kwargs)

Collect conf files and attempt to spawn the processes for this server

pid_files(**kwargs)

Get pid files for this server

Parameters

number – if supplied will only lookup the nth server

Returns

list of pid files

signal_children(sig, **kwargs)

Send a signal to child pids for this server

Parameters

sig – signal to send

Returns

a dict mapping pids (ints) to pid_files (paths)

signal_pids(sig, **kwargs)

Send a signal to pids for this server

Parameters

sig – signal to send

Returns

a dict mapping pids (ints) to pid_files (paths)

spawn(conf_file, once=False, wait=True, daemon=True, additional_args=None, **kwargs)

Launch a subprocess for this server.

Parameters
  • conf_file – path to conf_file to use as first arg

  • once – boolean, add once argument to command

  • wait – boolean, if true capture stdout with a pipe

  • daemon – boolean, if false ask server to log to console

  • additional_args – list of additional arguments to pass on the command line

Returns

the pid of the spawned process

status(pids=None, **kwargs)

Display status of server

Parameters
  • pids – if not supplied pids will be populated automatically

  • number – if supplied will only lookup the nth server

Returns

1 if server is not running, 0 otherwise

stop(**kwargs)

Send stop signals to pids for this server

Returns

a dict mapping pids (ints) to pid_files (paths)

wait(**kwargs)

wait on spawned procs to start

exception swift.common.manager.UnknownCommandError

Bases: Exception

swift.common.manager.command(func)

Decorator to declare which methods are accessible as commands, commands always return 1 or 0, where 0 should indicate success.

Parameters

func – function to make public

swift.common.manager.format_server_name(servername)

Formats server name as swift compatible server names E.g. swift-object-server

Parameters

servername – server name

Returns

swift compatible server name and its binary name

swift.common.manager.kill_group(pid, sig)

Send signal to process group

: param pid: process id : param sig: signal to send

swift.common.manager.safe_kill(pid, sig, name)

Send signal to process and check process name

: param pid: process id : param sig: signal to send : param name: name to ensure target process

swift.common.manager.setup_env()

Try to increase resource limits of the OS. Move PYTHON_EGG_CACHE to /tmp

swift.common.manager.verify_server(server)

Check whether the server is among swift servers or not, and also checks whether the server’s binaries are installed or not.

Parameters

server – name of the server

Returns

True, when the server name is valid and its binaries are found. False, otherwise.

swift.common.manager.watch_server_pids(server_pids, interval=1, **kwargs)

Monitor a collection of server pids yielding back those pids that aren’t responding to signals.

Parameters

server_pids – a dict, lists of pids [int,…] keyed on Server objects

MemCacheD

Why our own memcache client? By Michael Barton

python-memcached doesn’t use consistent hashing, so adding or removing a memcache server from the pool invalidates a huge percentage of cached items.

If you keep a pool of python-memcached client objects, each client object has its own connection to every memcached server, only one of which is ever in use. So you wind up with n * m open sockets and almost all of them idle. This client effectively has a pool for each server, so the number of backend connections is hopefully greatly reduced.

python-memcache uses pickle to store things, and there was already a huge stink about Swift using pickles in memcache (http://osvdb.org/show/osvdb/86581). That seemed sort of unfair, since nova and keystone and everyone else use pickles for memcache too, but it’s hidden behind a “standard” library. But changing would be a security regression at this point.

Also, pylibmc wouldn’t work for us because it needs to use python sockets in order to play nice with eventlet.

Lucid comes with memcached: v1.4.2. Protocol documentation for that version is at:

http://github.com/memcached/memcached/blob/1.4.2/doc/protocol.txt

class swift.common.memcached.MemcacheConnPool(server, size, connect_timeout, tls_context=None)

Bases: eventlet.pools.Pool

Connection pool for Memcache Connections

The server parameter can be a hostname, an IPv4 address, or an IPv6 address with an optional port. See swift.common.utils.parse_socket_string() for details.

create()

Generate a new pool item. In order for the pool to function, either this method must be overriden in a subclass or the pool must be constructed with the create argument. It accepts no arguments and returns a single instance of whatever thing the pool is supposed to contain.

In general, create() is called whenever the pool exceeds its previous high-water mark of concurrently-checked-out-items. In other words, in a new pool with min_size of 0, the very first call to get() will result in a call to create(). If the first caller calls put() before some other caller calls get(), then the first item will be returned, and create() will not be called a second time.

get()

Return an item from the pool, when one is available. This may cause the calling greenthread to block.

exception swift.common.memcached.MemcacheConnectionError

Bases: Exception

exception swift.common.memcached.MemcachePoolTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

class swift.common.memcached.MemcacheRing(servers, connect_timeout=0.3, io_timeout=2.0, pool_timeout=1.0, tries=3, max_conns=2, tls_context=None, logger=None, error_limit_count=10, error_limit_time=60, error_limit_duration=60, item_size_warning_threshold=- 1)

Bases: object

Simple, consistent-hashed memcache client.

decr(key, delta=1, time=0)

Decrements a key which has a numeric value by delta. Calls incr with -delta.

Parameters
  • key – key

  • delta – amount to subtract to the value of key (or set the value to 0 if the key is not found) will be cast to an int

  • time – the time to live

Returns

result of decrementing

Raises

MemcacheConnectionError

delete(key, server_key=None)

Deletes a key/value pair from memcache.

Parameters
  • key – key to be deleted

  • server_key – key to use in determining which server in the ring is used

get(key, raise_on_error=False)

Gets the object specified by key. It will also unserialize the object before returning if it is serialized in memcache with JSON.

Parameters
  • key – key

  • raise_on_error – if True, propagate Timeouts and other errors. By default, errors are treated as cache misses.

Returns

value of the key in memcache

get_multi(keys, server_key)

Gets multiple values from memcache for the given keys.

Parameters
  • keys – keys for values to be retrieved from memcache

  • server_key – key to use in determining which server in the ring is used

Returns

list of values

incr(key, delta=1, time=0)

Increments a key which has a numeric value by delta. If the key can’t be found, it’s added as delta or 0 if delta < 0. If passed a negative number, will use memcached’s decr. Returns the int stored in memcached Note: The data memcached stores as the result of incr/decr is an unsigned int. decr’s that result in a number below 0 are stored as 0.

Parameters
  • key – key

  • delta – amount to add to the value of key (or set as the value if the key is not found) will be cast to an int

  • time – the time to live

Returns

result of incrementing

Raises

MemcacheConnectionError

set(key, value, serialize=True, time=0, min_compress_len=0, raise_on_error=False)

Set a key/value pair in memcache

Parameters
  • key – key

  • value – value

  • serialize – if True, value is serialized with JSON before sending to memcache

  • time – the time to live

  • min_compress_len – minimum compress length, this parameter was added to keep the signature compatible with python-memcached interface. This implementation ignores it.

  • raise_on_error – if True, propagate Timeouts and other errors. By default, errors are ignored.

set_multi(mapping, server_key, serialize=True, time=0, min_compress_len=0)

Sets multiple key/value pairs in memcache.

Parameters
  • mapping – dictionary of keys and values to be set in memcache

  • server_key – key to use in determining which server in the ring is used

  • serialize – if True, value is serialized with JSON before sending to memcache.

  • time – the time to live

Min_compress_len

minimum compress length, this parameter was added to keep the signature compatible with python-memcached interface. This implementation ignores it

swift.common.memcached.load_memcache(conf, logger)

Build a MemcacheRing object from the given config. It will also use the passed in logger.

Parameters
  • conf – a dict, the config options

  • logger – a logger

swift.common.memcached.sanitize_timeout(timeout)

Sanitize a timeout value to use an absolute expiration time if the delta is greater than 30 days (in seconds). Note that the memcached server translates negative values to mean a delta of 30 days in seconds (and 1 additional second), client beware.

Middleware Registry

swift.common.registry.get_sensitive_headers()

Returns the set of registered sensitive headers.

Used by swift.common.middleware.proxy_logging to perform redactions prior to logging.

swift.common.registry.get_sensitive_params()

Returns the set of registered sensitive query parameters.

Used by swift.common.middleware.proxy_logging to perform redactions prior to logging.

swift.common.registry.get_swift_info(admin=False, disallowed_sections=None)

Returns information about the swift cluster that has been previously registered with the register_swift_info call.

Parameters
  • admin – boolean value, if True will additionally return an ‘admin’ section with information previously registered as admin info.

  • disallowed_sections – list of section names to be withheld from the information returned.

Returns

dictionary of information about the swift cluster.

swift.common.registry.register_sensitive_header(header)

Register a header as being “sensitive”.

Sensitive headers are automatically redacted when logging. See the reveal_sensitive_prefix option in the proxy-server sample config for more information.

Parameters

header – The (case-insensitive) header name which, if present, may contain sensitive information. Examples include X-Auth-Token and (if s3api is enabled) Authorization. Limited to ASCII characters.

swift.common.registry.register_sensitive_param(query_param)

Register a query parameter as being “sensitive”.

Sensitive query parameters are automatically redacted when logging. See the reveal_sensitive_prefix option in the proxy-server sample config for more information.

Parameters

query_param – The (case-sensitive) query parameter name which, if present, may contain sensitive information. Examples include temp_url_signature and (if s3api is enabled) X-Amz-Signature. Limited to ASCII characters.

swift.common.registry.register_swift_info(name='swift', admin=False, **kwargs)

Registers information about the swift cluster to be retrieved with calls to get_swift_info.

NOTE: Do not use “.” in the param: name or any keys in kwargs. “.” is used

in the disallowed_sections to remove unwanted keys from /info.

Parameters
  • name – string, the section name to place the information under.

  • admin – boolean, if True, information will be registered to an admin section which can optionally be withheld when requesting the information.

  • kwargs – key value arguments representing the information to be added.

Raises

ValueError – if name or any of the keys in kwargs has “.” in it

Request Helpers

Miscellaneous utility functions for use in generating responses.

Why not swift.common.utils, you ask? Because this way we can import things from swob in here without creating circular imports.

class swift.common.request_helpers.SegmentedIterable(req, app, listing_iter, max_get_time, logger, ua_suffix, swift_source, name='<not specified>', response_body_length=None)

Bases: object

Iterable that returns the object contents for a large object.

Parameters
  • req – original request object

  • app – WSGI application from which segments will come

  • listing_iter

    iterable yielding the object segments to fetch, along with the byte subranges to fetch, in the form of a 5-tuple (object-path, object-etag, object-size, first-byte, last-byte).

    If object-etag is None, no MD5 verification will be done.

    If object-size is None, no length verification will be done.

    If first-byte and last-byte are None, then the entire object will be fetched.

  • max_get_time – maximum permitted duration of a GET request (seconds)

  • logger – logger object

  • swift_source – value of swift.source in subrequest environ (just for logging)

  • ua_suffix – string to append to user-agent.

  • name – name of manifest (used in logging only)

  • response_body_length – optional response body length for the response being sent to the client.

app_iter_range(*a, **kw)

swob.Response will only respond with a 206 status in certain cases; one of those is if the body iterator responds to .app_iter_range().

However, this object (or really, its listing iter) is smart enough to handle the range stuff internally, so we just no-op this out for swob.

app_iter_ranges(ranges, content_type, boundary, content_size)

This method assumes that iter(self) yields all the data bytes that go into the response, but none of the MIME stuff. For example, if the response will contain three MIME docs with data “abcd”, “efgh”, and “ijkl”, then iter(self) will give out the bytes “abcdefghijkl”.

This method inserts the MIME stuff around the data bytes.

close()

Called when the client disconnect. Ensure that the connection to the backend server is closed.

validate_first_segment()

Start fetching object data to ensure that the first segment (if any) is valid. This is to catch cases like “first segment is missing” or “first segment’s etag doesn’t match manifest”.

Note: this does not validate that you have any segments. A zero-segment large object is not erroneous; it is just empty.

swift.common.request_helpers.check_path_header(req, name, length, error_msg)

Validate that the value of path-like header is well formatted. We assume the caller ensures that specific header is present in req.headers.

Parameters
  • req – HTTP request object

  • name – header name

  • length – length of path segment check

  • error_msg – error message for client

Returns

A tuple with path parts according to length

Raise

HTTPPreconditionFailed if header value is not well formatted.

swift.common.request_helpers.constrain_req_limit(req, constrained_limit)
swift.common.request_helpers.copy_header_subset(from_r, to_r, condition)

Will copy desired subset of headers from from_r to to_r.

Parameters
  • from_r – a swob Request or Response

  • to_r – a swob Request or Response

  • condition – a function that will be passed the header key as a single argument and should return True if the header is to be copied.

swift.common.request_helpers.get_container_update_override_key(key)

Returns the full X-Object-Sysmeta-Container-Update-Override-* header key.

Parameters

key – the key you want to override in the container update

Returns

the full header key

swift.common.request_helpers.get_ip_port(node, headers)

Get the ip address and port that should be used for the given node. The normal ip address and port are returned unless the node or headers indicate that the replication ip address and port should be used.

If the headers dict has an item with key x-backend-use-replication-network and a truthy value then the replication ip address and port are returned. Otherwise if the node dict has an item with key use_replication and truthy value then the replication ip address and port are returned. Otherwise the normal ip address and port are returned.

Parameters
  • node – a dict describing a node

  • headers – a dict of headers

Returns

a tuple of (ip address, port)

swift.common.request_helpers.get_name_and_placement(request, minsegs=1, maxsegs=None, rest_with_last=False)

Utility function to split and validate the request path and storage policy. The storage policy index is extracted from the headers of the request and converted to a StoragePolicy instance. The remaining args are passed through to split_and_validate_path().

Returns

a list, result of split_and_validate_path() with the BaseStoragePolicy instance appended on the end

Raises

HTTPServiceUnavailable – if the path is invalid or no policy exists with the extracted policy_index.

swift.common.request_helpers.get_object_transient_sysmeta(key)

Returns the Object Transient System Metadata header for key. The Object Transient System Metadata namespace will be persisted by backend object servers. These headers are treated in the same way as object user metadata i.e. all headers in this namespace will be replaced on every POST request.

Parameters

key – metadata key

Returns

the entire object transient system metadata header for key

swift.common.request_helpers.get_param(req, name, default=None)

Get a parameter from an HTTP request ensuring proper handling UTF-8 encoding.

Parameters
  • req – request object

  • name – parameter name

  • default – result to return if the parameter is not found

Returns

HTTP request parameter value, as a native string (in py2, as UTF-8 encoded str, not unicode object)

Raises

HTTPBadRequest – if param not valid UTF-8 byte sequence

swift.common.request_helpers.get_reserved_name(*parts)

Generate a valid reserved name that joins the component parts.

Returns

a string

swift.common.request_helpers.get_sys_meta_prefix(server_type)

Returns the prefix for system metadata headers for given server type.

This prefix defines the namespace for headers that will be persisted by backend servers.

Parameters

server_type – type of backend server i.e. [account|container|object]

Returns

prefix string for server type’s system metadata headers

swift.common.request_helpers.get_user_meta_prefix(server_type)

Returns the prefix for user metadata headers for given server type.

This prefix defines the namespace for headers that will be persisted by backend servers.

Parameters

server_type – type of backend server i.e. [account|container|object]

Returns

prefix string for server type’s user metadata headers

swift.common.request_helpers.http_response_to_document_iters(response, read_chunk_size=4096)

Takes a successful object-GET HTTP response and turns it into an iterator of (first-byte, last-byte, length, headers, body-file) 5-tuples.

The response must either be a 200 or a 206; if you feed in a 204 or something similar, this probably won’t work.

Parameters

response – HTTP response, like from bufferedhttp.http_connect(), not a swob.Response.

swift.common.request_helpers.is_object_transient_sysmeta(key)

Tests if a header key starts with and is longer than the prefix for object transient system metadata.

Parameters

key – header key

Returns

True if the key satisfies the test, False otherwise

swift.common.request_helpers.is_sys_meta(server_type, key)

Tests if a header key starts with and is longer than the system metadata prefix for given server type.

Parameters
  • server_type – type of backend server i.e. [account|container|object]

  • key – header key

Returns

True if the key satisfies the test, False otherwise

swift.common.request_helpers.is_sys_or_user_meta(server_type, key)

Tests if a header key starts with and is longer than the user or system metadata prefix for given server type.

Parameters
  • server_type – type of backend server i.e. [account|container|object]

  • key – header key

Returns

True if the key satisfies the test, False otherwise

swift.common.request_helpers.is_use_replication_network(headers=None)

Determine if replication network should be used.

Parameters

headers – a dict of headers

Returns

the value of the x-backend-use-replication-network item from headers. If no headers are given or the item is not found then False is returned.

swift.common.request_helpers.is_user_meta(server_type, key)

Tests if a header key starts with and is longer than the user metadata prefix for given server type.

Parameters
  • server_type – type of backend server i.e. [account|container|object]

  • key – header key

Returns

True if the key satisfies the test, False otherwise

swift.common.request_helpers.remove_items(headers, condition)

Removes items from a dict whose keys satisfy the given condition.

Parameters
  • headers – a dict of headers

  • condition – a function that will be passed the header key as a single argument and should return True if the header is to be removed.

Returns

a dict, possibly empty, of headers that have been removed

swift.common.request_helpers.resolve_etag_is_at_header(req, metadata)

Helper function to resolve an alternative etag value that may be stored in metadata under an alternate name.

The value of the request’s X-Backend-Etag-Is-At header (if it exists) is a comma separated list of alternate names in the metadata at which an alternate etag value may be found. This list is processed in order until an alternate etag is found.

The left most value in X-Backend-Etag-Is-At will have been set by the left most middleware, or if no middleware, by ECObjectController, if an EC policy is in use. The left most middleware is assumed to be the authority on what the etag value of the object content is.

The resolver will work from left to right in the list until it finds a value that is a name in the given metadata. So the left most wins, IF it exists in the metadata.

By way of example, assume the encrypter middleware is installed. If an object is not encrypted then the resolver will not find the encrypter middleware’s alternate etag sysmeta (X-Object-Sysmeta-Crypto-Etag) but will then find the EC alternate etag (if EC policy). But if the object is encrypted then X-Object-Sysmeta-Crypto-Etag is found and used, which is correct because it should be preferred over X-Object-Sysmeta-Ec-Etag.

Parameters
  • req – a swob Request

  • metadata – a dict containing object metadata

Returns

an alternate etag value if any is found, otherwise None

swift.common.request_helpers.split_and_validate_path(request, minsegs=1, maxsegs=None, rest_with_last=False)

Utility function to split and validate the request path.

Returns

result of split_path() if everything’s okay, as native strings

Raises

HTTPBadRequest – if something’s not okay

swift.common.request_helpers.split_reserved_name(name)

Separate a valid reserved name into the component parts.

Returns

a list of strings

swift.common.request_helpers.strip_object_transient_sysmeta_prefix(key)

Removes the object transient system metadata prefix from the start of a header key.

Parameters

key – header key

Returns

stripped header key

swift.common.request_helpers.strip_sys_meta_prefix(server_type, key)

Removes the system metadata prefix for a given server type from the start of a header key.

Parameters
  • server_type – type of backend server i.e. [account|container|object]

  • key – header key

Returns

stripped header key

swift.common.request_helpers.strip_user_meta_prefix(server_type, key)

Removes the user metadata prefix for a given server type from the start of a header key.

Parameters
  • server_type – type of backend server i.e. [account|container|object]

  • key – header key

Returns

stripped header key

swift.common.request_helpers.update_etag_is_at_header(req, name)

Helper function to update an X-Backend-Etag-Is-At header whose value is a list of alternative header names at which the actual object etag may be found. This informs the object server where to look for the actual object etag when processing conditional requests.

Since the proxy server and/or middleware may set alternative etag header names, the value of X-Backend-Etag-Is-At is a comma separated list which the object server inspects in order until it finds an etag value.

Parameters
  • req – a swob Request

  • name – name of a sysmeta where alternative etag may be found

swift.common.request_helpers.update_ignore_range_header(req, name)

Helper function to update an X-Backend-Ignore-Range-If-Metadata-Present header whose value is a list of header names which, if any are present on an object, mean the object server should respond with a 200 instead of a 206 or 416.

Parameters
  • req – a swob Request

  • name – name of a header which, if found, indicates the proxy will want the whole object

swift.common.request_helpers.validate_container_params(req)
swift.common.request_helpers.validate_internal_account(account)

Validate internal account name.

Raises

HTTPBadRequest

swift.common.request_helpers.validate_internal_container(account, container)

Validate internal account and container names.

Raises

HTTPBadRequest

swift.common.request_helpers.validate_internal_obj(account, container, obj)

Validate internal account, container and object names.

Raises

HTTPBadRequest

swift.common.request_helpers.validate_params(req, names)

Get list of parameters from an HTTP request, validating the encoding of each parameter.

Parameters
  • req – request object

  • names – parameter names

Returns

a dict mapping parameter names to values for each name that appears in the request parameters

Raises

HTTPBadRequest – if any parameter value is not a valid UTF-8 byte sequence

Swob

Implementation of WSGI Request and Response objects.

This library has a very similar API to Webob. It wraps WSGI request environments and response values into objects that are more friendly to interact with.

Why Swob and not just use WebOb? By Michael Barton

We used webob for years. The main problem was that the interface wasn’t stable. For a while, each of our several test suites required a slightly different version of webob to run, and none of them worked with the then-current version. It was a huge headache, so we just scrapped it.

This is kind of a ton of code, but it’s also been a huge relief to not have to scramble to add a bunch of code branches all over the place to keep Swift working every time webob decides some interface needs to change.

class swift.common.swob.Accept(headerval)

Bases: object

Wraps a Request’s Accept header as a friendly object.

Parameters

headerval – value of the header as a str

best_match(options)

Returns the item from “options” that best matches the accept header. Returns None if no available options are acceptable to the client.

Parameters

options – a list of content-types the server can respond with

Raises

ValueError – if the header is malformed

exception swift.common.swob.HTTPException(*args, **kwargs)

Bases: swift.common.swob.Response, Exception

class swift.common.swob.HeaderEnvironProxy(environ)

Bases: collections.abc.MutableMapping

A dict-like object that proxies requests to a wsgi environ, rewriting header keys to environ keys.

For example, headers[‘Content-Range’] sets and gets the value of headers.environ[‘HTTP_CONTENT_RANGE’]

keys() a set-like object providing a view on D's keys
class swift.common.swob.Match(headerval)

Bases: object

Wraps a Request’s If-[None-]Match header as a friendly object.

Parameters

headerval – value of the header as a str

class swift.common.swob.Range(headerval)

Bases: object

Wraps a Request’s Range header as a friendly object. After initialization, “range.ranges” is populated with a list of (start, end) tuples denoting the requested ranges.

If there were any syntactically-invalid byte-range-spec values, the constructor will raise a ValueError, per the relevant RFC:

“The recipient of a byte-range-set that includes one or more syntactically invalid byte-range-spec values MUST ignore the header field that includes that byte-range-set.”

According to the RFC 2616 specification, the following cases will be all considered as syntactically invalid, thus, a ValueError is thrown so that the range header will be ignored. If the range value contains at least one of the following cases, the entire range is considered invalid, ValueError will be thrown so that the header will be ignored.

  1. value not starts with bytes=

  2. range value start is greater than the end, eg. bytes=5-3

  3. range does not have start or end, eg. bytes=-

  4. range does not have hyphen, eg. bytes=45

  5. range value is non numeric

  6. any combination of the above

Every syntactically valid range will be added into the ranges list even when some of the ranges may not be satisfied by underlying content.

Parameters

headerval – value of the header as a str

ranges_for_length(length)

This method is used to return multiple ranges for a given length which should represent the length of the underlying content. The constructor method __init__ made sure that any range in ranges list is syntactically valid. So if length is None or size of the ranges is zero, then the Range header should be ignored which will eventually make the response to be 200.

If an empty list is returned by this method, it indicates that there are unsatisfiable ranges found in the Range header, 416 will be returned.

if a returned list has at least one element, the list indicates that there is at least one range valid and the server should serve the request with a 206 status code.

The start value of each range represents the starting position in the content, the end value represents the ending position. This method purposely adds 1 to the end number because the spec defines the Range to be inclusive.

The Range spec can be found at the following link: http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.35.1

Parameters

length – length of the underlying content

class swift.common.swob.Request(environ)

Bases: object

WSGI Request object.

property accept

Retrieve and set the accept property in the WSGI environ, as a Accept object

property acl

Get and set the swob.ACL property in the WSGI environment

classmethod blank(path, environ=None, headers=None, body=None, **kwargs)

Create a new request object with the given parameters, and an environment otherwise filled in with non-surprising default values.

Parameters
  • path – encoded, parsed, and unquoted into PATH_INFO

  • environ – WSGI environ dictionary

  • headers – HTTP headers

  • body – stuffed in a WsgiBytesIO and hung on wsgi.input

  • kwargs – any environ key with an property setter

property body

Get and set the request body str

property body_file

Get and set the wsgi.input property in the WSGI environment

call_application(application)

Calls the application with this request’s environment. Returns the status, headers, and app_iter for the response as a tuple.

Parameters

application – the WSGI application to call

property content_length

Retrieve and set the content-length header as an int

copy_get()

Makes a copy of the request, converting it to a GET.

ensure_x_timestamp()

Similar to timestamp, but the X-Timestamp header will be set if not present.

Raises

HTTPBadRequest – if X-Timestamp is already set but not a valid Timestamp

Returns

the request’s X-Timestamp header, as a Timestamp

get_response(application)

Calls the application with this request’s environment. Returns a Response object that wraps up the application’s result.

Parameters

application – the WSGI application to call

property host

Get and set the HTTP_HOST property in the WSGI environment

property host_url

Get url for request/response up to path

property if_match

Retrieve and set the if-match property in the WSGI environ, as a Match object

property if_modified_since

Retrieve and set the if-modified-since header as a datetime, set it with a datetime, int, or str

property if_none_match

Retrieve and set the if-none-match property in the WSGI environ, as a Match object

property if_unmodified_since

Retrieve and set the if-unmodified-since header as a datetime, set it with a datetime, int, or str

message_length()

Properly determine the message length for this request. It will return an integer if the headers explicitly contain the message length, or None if the headers don’t contain a length. The ValueError exception will be raised if the headers are invalid.

Raises
  • ValueError – if either transfer-encoding or content-length headers have bad values

  • AttributeError – if the last value of the transfer-encoding header is not “chunked”

property method

Get and set the REQUEST_METHOD property in the WSGI environment

property params

Provides QUERY_STRING parameters as a dictionary

property path

Provides the full path of the request, excluding the QUERY_STRING

property path_info

Get and set the PATH_INFO property in the WSGI environment

path_info_pop()

Takes one path portion (delineated by slashes) from the path_info, and appends it to the script_name. Returns the path segment.

property path_qs

The path of the request, without host but with query string.

property query_string

Get and set the QUERY_STRING property in the WSGI environment

property range

Retrieve and set the range property in the WSGI environ, as a Range object

property referer

Get and set the HTTP_REFERER property in the WSGI environment

property referrer

Get and set the HTTP_REFERER property in the WSGI environment

property remote_addr

Get and set the REMOTE_ADDR property in the WSGI environment

property remote_user

Get and set the REMOTE_USER property in the WSGI environment

property script_name

Get and set the SCRIPT_NAME property in the WSGI environment

split_path(minsegs=1, maxsegs=None, rest_with_last=False)

Validate and split the Request’s path.

Examples:

['a'] = split_path('/a')
['a', None] = split_path('/a', 1, 2)
['a', 'c'] = split_path('/a/c', 1, 2)
['a', 'c', 'o/r'] = split_path('/a/c/o/r', 1, 3, True)
Parameters
  • minsegs – Minimum number of segments to be extracted

  • maxsegs – Maximum number of segments to be extracted

  • rest_with_last – If True, trailing data will be returned as part of last segment. If False, and there is trailing data, raises ValueError.

Returns

list of segments with a length of maxsegs (non-existent segments will return as None)

Raises

ValueError – if given an invalid path

property str_params

Provides QUERY_STRING parameters as a dictionary

property swift_entity_path

Provides the (native string) account/container/object path, sans API version.

This can be useful when constructing a path to send to a backend server, as that path will need everything after the “/v1”.

property timestamp

Provides HTTP_X_TIMESTAMP as a Timestamp

property url

Provides the full url of the request

property user_agent

Get and set the HTTP_USER_AGENT property in the WSGI environment

class swift.common.swob.Response(body=None, status=200, headers=None, app_iter=None, request=None, conditional_response=False, conditional_etag=None, **kw)

Bases: object

WSGI Response object.

__call__(env, start_response)

Respond to the WSGI request.

Warning

This will translate any relative Location header value to an absolute URL using the WSGI environment’s HOST_URL as a prefix, as RFC 2616 specifies.

However, it is quite common to use relative redirects, especially when it is difficult to know the exact HOST_URL the browser would have used when behind several CNAMEs, CDN services, etc. All modern browsers support relative redirects.

To skip over RFC enforcement of the Location header value, you may set env['swift.leave_relative_location'] = True in the WSGI environment.

absolute_location()

Attempt to construct an absolute location.

property accept_ranges

Retrieve and set the accept-ranges header

property app_iter

Retrieve and set the response app_iter

property body

Retrieve and set the Response body str

property charset

Retrieve and set the response charset

property conditional_etag

The conditional_etag keyword argument for Response will allow the conditional match value of a If-Match request to be compared to a non-standard value.

This is available for Storage Policies that do not store the client object data verbatim on the storage nodes, but still need support conditional requests.

It’s most effectively used with X-Backend-Etag-Is-At which would define the additional Metadata key(s) where the original ETag of the clear-form client request data may be found.

property content_length

Retrieve and set the content-length header as an int

property content_range

Retrieve and set the content-range header

property content_type

Retrieve and set the response Content-Type header

property etag

Retrieve and set the response Etag header

fix_conditional_response()

You may call this once you have set the content_length to the whole object length and body or app_iter to reset the content_length properties on the request.

It is ok to not call this method, the conditional response will be maintained for you when you __call__ the response.

property host_url

Get url for request/response up to path

property last_modified

Retrieve and set the last-modified header as a datetime, set it with a datetime, int, or str

property location

Retrieve and set the location header

property status

Retrieve and set the Response status, e.g. ‘200 OK’

www_authenticate()

Construct a suitable value for WWW-Authenticate response header

If we have a request and a valid-looking path, the realm is the account; otherwise we set it to ‘unknown’.

class swift.common.swob.StatusMap

Bases: object

A dict-like object that returns HTTPException subclasses/factory functions where the given key is the status code.

class swift.common.swob.WsgiBytesIO(initial_bytes=b'')

Bases: _io.BytesIO

This class adds support for the additional wsgi.input methods defined on eventlet.wsgi.Input to the BytesIO class which would otherwise be a fine stand-in for the file-like object in the WSGI environment.

swift.common.swob.wsgify(func)

A decorator for translating functions which take a swob Request object and return a Response object into WSGI callables. Also catches any raised HTTPExceptions and treats them as a returned Response.

Utils

Miscellaneous utility functions for use with Swift.

swift.common.utils.ATTRIBUTES_RE = re.compile('(\\w+)=(".*?"|[^";]+)(; ?|$)')

Regular expression to match form attributes.

class swift.common.utils.CloseableChain(*iterables)

Bases: object

Like itertools.chain, but with a close method that will attempt to invoke its sub-iterators’ close methods, if any.

class swift.common.utils.ContextPool(size=1000)

Bases: eventlet.greenpool.GreenPool

GreenPool subclassed to kill its coros when it gets gc’ed

class swift.common.utils.Everything

Bases: object

A container that contains everything. If “e” is an instance of Everything, then “x in e” is true for all x.

class swift.common.utils.GreenAsyncPile(size_or_pool)

Bases: object

Runs jobs in a pool of green threads, and the results can be retrieved by using this object as an iterator.

This is very similar in principle to eventlet.GreenPile, except it returns results as they become available rather than in the order they were launched.

Correlating results with jobs (if necessary) is left to the caller.

spawn(func, *args, **kwargs)

Spawn a job in a green thread on the pile.

waitall(timeout)

Wait timeout seconds for any results to come in.

Parameters

timeout – seconds to wait for results

Returns

list of results accrued in that time

waitfirst(timeout)

Wait up to timeout seconds for first result to come in.

Parameters

timeout – seconds to wait for results

Returns

first item to come back, or None

exception swift.common.utils.GreenAsyncPileWaitallTimeout(seconds=None, exception=None)

Bases: eventlet.timeout.Timeout

class swift.common.utils.GreenthreadSafeIterator(unsafe_iterable)

Bases: object

Wrap an iterator to ensure that only one greenthread is inside its next() method at a time.

This is useful if an iterator’s next() method may perform network IO, as that may trigger a greenthread context switch (aka trampoline), which can give another greenthread a chance to call next(). At that point, you get an error like “ValueError: generator already executing”. By wrapping calls to next() with a mutex, we avoid that error.

class swift.common.utils.InputProxy(wsgi_input)

Bases: object

File-like object that counts bytes read. To be swapped in for wsgi.input for accounting purposes.

read(*args, **kwargs)

Pass read request to the underlying file-like object and add bytes read to total.

readline(*args, **kwargs)

Pass readline request to the underlying file-like object and add bytes read to total.

exception swift.common.utils.InvalidHashPathConfigError

Bases: ValueError

class swift.common.utils.LRUCache(maxsize=1000, maxtime=3600)

Bases: object

Decorator for size/time bound memoization that evicts the least recently used members.

class swift.common.utils.LogAdapter(logger, server)

Bases: logging.LoggerAdapter, object

A Logger like object which performs some reformatting on calls to exception(). Can be used to store a threadlocal transaction id and client ip.

exception(msg, *args, **kwargs)

Delegate an exception call to the underlying logger.

getEffectiveLevel()

Get the effective level for the underlying logger.

notice(msg, *args, **kwargs)

Convenience function for syslog priority LOG_NOTICE. The python logging lvl is set to 25, just above info. SysLogHandler is monkey patched to map this log lvl to the LOG_NOTICE syslog priority.

process(msg, kwargs)

Add extra info to message

set_statsd_prefix(prefix)

This method is deprecated. Callers should use the statsd_tail_prefix argument of get_logger when instantiating a logger.

The StatsD client prefix defaults to the “name” of the logger. This method may override that default with a specific value. Currently used in the proxy-server to differentiate the Account, Container, and Object controllers.

statsd_delegate()

Factory to create methods which delegate to methods on self.logger.statsd_client (an instance of StatsdClient). The created methods conditionally delegate to a method whose name is given in ‘statsd_func_name’. The created delegate methods are a no-op when StatsD logging is not configured.

Parameters

statsd_func_name – the name of a method on StatsdClient.

class swift.common.utils.LogLevelFilter(level=10)

Bases: object

Drop messages for the logger based on level.

This is useful when dependencies log too much information.

Parameters

level – All messages at or below this level are dropped (DEBUG < INFO < WARN < ERROR < CRITICAL|FATAL) Default: DEBUG

class swift.common.utils.MetricsPrefixLoggerAdapter(logger, extra, metric_prefix)

Bases: swift.common.utils.SwiftLoggerAdapter

Adds a prefix to all Statsd metrics’ names.

swift.common.utils.NR_ioprio_set()

Give __NR_ioprio_set value for your system.

class swift.common.utils.NicerInterpolation

Bases: configparser.BasicInterpolation

class swift.common.utils.NoopMutex

Bases: object

“Mutex” that doesn’t lock anything.

We only allow our syslog logging to be configured via UDS or UDP, neither of which have the message-interleaving trouble you’d expect from TCP or file handlers.

class swift.common.utils.NullLogger

Bases: object

A no-op logger for eventlet wsgi.

class swift.common.utils.OverrideOptions(devices, partitions, policies)

Bases: tuple

devices

Alias for field number 0

partitions

Alias for field number 1

policies

Alias for field number 2

class swift.common.utils.PipeMutex

Bases: object

Mutex using a pipe. Works across both greenlets and real threads, even at the same time.

acquire(blocking=True)

Acquire the mutex.

If called with blocking=False, returns True if the mutex was acquired and False if it wasn’t. Otherwise, blocks until the mutex is acquired and returns True.

This lock is recursive; the same greenthread may acquire it as many times as it wants to, though it must then release it that many times too.

close()

Close the mutex. This releases its file descriptors.

You can’t use a mutex after it’s been closed.

release()

Release the mutex.

class swift.common.utils.PrefixLoggerAdapter(logger, extra=None)

Bases: swift.common.utils.SwiftLoggerAdapter

Adds an optional prefix to all its log messages. When the prefix has not been set, messages are unchanged.

exception(msg, *a, **kw)

Delegate an exception call to the underlying logger.

process(msg, kwargs)

Process the logging message and keyword arguments passed in to a logging call to insert contextual information. You can either manipulate the message itself, the keyword args or both. Return the message and kwargs modified (or not) to suit your needs.

Normally, you’ll only need to override this one method in a LoggerAdapter subclass for your specific needs.

class swift.common.utils.RateLimitedIterator(iterable, elements_per_second, limit_after=0, ratelimit_if=<function RateLimitedIterator.<lambda>>)

Bases: object

Wrap an iterator to only yield elements at a rate of N per second.

Parameters
  • iterable – iterable to wrap

  • elements_per_second – the rate at which to yield elements

  • limit_after – rate limiting kicks in only after yielding this many elements; default is 0 (rate limit immediately)

class swift.common.utils.ShardName(account, root_container, parent_container_hash, timestamp, index)

Bases: object

Encapsulates the components of a shard name.

Instances of this class would typically be constructed via the create() or parse() class methods.

Shard names have the form:

<account>/<root_container>-<parent_container_hash>-<timestamp>-<index>

Note: some instances of ShardRange have names that will NOT parse as a ShardName; e.g. a root container’s own shard range will have a name format of <account>/<root_container> which will raise ValueError if passed to parse.

classmethod create(account, root_container, parent_container, timestamp, index)

Create an instance of ShardName.

Parameters
  • account – the hidden internal account to which the shard container belongs.

  • root_container – the name of the root container for the shard.

  • parent_container – the name of the parent container for the shard; for initial first generation shards this should be the same as root_container; for shards of shards this should be the name of the sharding shard container.

  • timestamp – an instance of Timestamp

  • index – a unique index that will distinguish the path from any other path generated using the same combination of account, root_container, parent_container and timestamp.

Returns

an instance of ShardName.

Raises

ValueError – if any argument is None

classmethod hash_container_name(container_name)

Calculates the hash of a container name.

Parameters

container_name – name to be hashed.

Returns

the hexdigest of the md5 hash of container_name.

Raises

ValueError – if container_name is None.

classmethod parse(name)

Parse name to an instance of ShardName.

Parameters

name – a shard name which should have the form: <account>/ <root_container>-<parent_container_hash>-<timestamp>-<index>

Returns

an instance of ShardName.

Raises

ValueError – if name is not a valid shard name.

class swift.common.utils.ShardRange(name, timestamp, lower=MinBound, upper=MaxBound, object_count=0, bytes_used=0, meta_timestamp=None, deleted=False, state=None, state_timestamp=None, epoch=None, reported=False, tombstones=- 1)

Bases: object

A ShardRange encapsulates sharding state related to a container including lower and upper bounds that define the object namespace for which the container is responsible.

Shard ranges may be persisted in a container database. Timestamps associated with subsets of the shard range attributes are used to resolve conflicts when a shard range needs to be merged with an existing shard range record and the most recent version of an attribute should be persisted.

Parameters
  • name – the name of the shard range; this should take the form of a path to a container i.e. <account_name>/<container_name>.

  • timestamp – a timestamp that represents the time at which the shard range’s lower, upper or deleted attributes were last modified.

  • lower – the lower bound of object names contained in the shard range; the lower bound is not included in the shard range namespace.

  • upper – the upper bound of object names contained in the shard range; the upper bound is included in the shard range namespace.

  • object_count – the number of objects in the shard range; defaults to zero.

  • bytes_used – the number of bytes in the shard range; defaults to zero.

  • meta_timestamp – a timestamp that represents the time at which the shard range’s object_count and bytes_used were last updated; defaults to the value of timestamp.

  • deleted – a boolean; if True the shard range is considered to be deleted.

  • state – the state; must be one of ShardRange.STATES; defaults to CREATED.

  • state_timestamp – a timestamp that represents the time at which state was forced to its current value; defaults to the value of timestamp. This timestamp is typically not updated with every change of state because in general conflicts in state attributes are resolved by choosing the larger state value. However, when this rule does not apply, for example when changing state from SHARDED to ACTIVE, the state_timestamp may be advanced so that the new state value is preferred over any older state value.

  • epoch – optional epoch timestamp which represents the time at which sharding was enabled for a container.

  • reported – optional indicator that this shard and its stats have been reported to the root container.

  • tombstones – the number of tombstones in the shard range; defaults to -1 to indicate that the value is unknown.

class MaxBound

Bases: swift.common.utils.ShardRangeOuterBound

class MinBound

Bases: swift.common.utils.ShardRangeOuterBound

copy(timestamp=None, **kwargs)

Creates a copy of the ShardRange.

Parameters

timestamp – (optional) If given, the returned ShardRange will have all of its timestamps set to this value. Otherwise the returned ShardRange will have the original timestamps.

Returns

an instance of ShardRange

entire_namespace()

Returns True if the ShardRange includes the entire namespace, False otherwise.

expand(donors)

Expands the bounds as necessary to match the minimum and maximum bounds of the given donors.

Parameters

donors – A list of ShardRange

Returns

True if the bounds have been modified, False otherwise.

find_ancestors(shard_ranges)

Find this shard range’s ancestor ranges in the given shard_ranges.

This method makes a best-effort attempt to identify this shard range’s parent shard range, the parent’s parent, etc., up to and including the root shard range. It is only possible to directly identify the parent of a particular shard range, so the search is recursive; if any member of the ancestry is not found then the search ends and older ancestors that may be in the list are not identified. The root shard range, however, will always be identified if it is present in the list.

For example, given a list that contains parent, grandparent, great-great-grandparent and root shard ranges, but is missing the great-grandparent shard range, only the parent, grand-parent and root shard ranges will be identified.

Parameters

shard_ranges – a list of instances of ShardRange

Returns

a list of instances of ShardRange containing items in the given shard_ranges that can be identified as ancestors of this shard range. The list may not be complete if there are gaps in the ancestry, but is guaranteed to contain at least the parent and root shard ranges if they are present.

find_root(shard_ranges)

Find this shard range’s root shard range in the given shard_ranges.

Parameters

shard_ranges – a list of instances of ShardRange

Returns

this shard range’s root shard range if it is found in the list, otherwise None.

classmethod from_dict(params)

Return an instance constructed using the given dict of params. This method is deliberately less flexible than the class __init__() method and requires all of the __init__() args to be given in the dict of params.

Parameters

params – a dict of parameters

Returns

an instance of this class

includes(other)

Returns True if this namespace includes the whole of the other namespace, False otherwise.

Parameters

other – an instance of ShardRange

increment_meta(object_count, bytes_used)

Increment the object stats metadata by the given values and update the meta_timestamp to the current time.

Parameters
  • object_count – should be an integer

  • bytes_used – should be an integer

Raises

ValueError – if object_count or bytes_used cannot be cast to an int.

is_child_of(parent)

Test if this shard range is a child of another shard range. The parent-child relationship is inferred from the names of the shard ranges. This method is limited to work only within the scope of the same user-facing account (with and without shard prefix).

Parameters

parent – an instance of ShardRange.

Returns

True if parent is the parent of this shard range, False otherwise, assuming that they are within the same account.

classmethod make_path(shards_account, root_container, parent_container, timestamp, index)

Returns a path for a shard container that is valid to use as a name when constructing a ShardRange.

Parameters
  • shards_account – the hidden internal account to which the shard container belongs.

  • root_container – the name of the root container for the shard.

  • parent_container – the name of the parent container for the shard; for initial first generation shards this should be the same as root_container; for shards of shards this should be the name of the sharding shard container.

  • timestamp – an instance of Timestamp

  • index – a unique index that will distinguish the path from any other path generated using the same combination of shards_account, root_container, parent_container and timestamp.

Returns

a string of the form <account_name>/<container_name>

overlaps(other)

Returns True if the ShardRange namespace overlaps with the other ShardRange’s namespace.

Parameters

other – an instance of ShardRange

classmethod resolve_state(state)

Given a value that may be either the name or the number of a state return a tuple of (state number, state name).

Parameters

state – Either a string state name or an integer state number.

Returns

A tuple (state number, state name)

Raises

ValueError – if state is neither a valid state name nor a valid state number.

property row_count

Returns the total number of rows in the shard range i.e. the sum of objects and tombstones.

Returns

the row count

set_deleted(timestamp=None)

Mark the shard range deleted and set timestamp to the current time.

Parameters

timestamp – optional timestamp to set; if not given the current time will be set.

Returns

True if the deleted attribute or timestamp was changed, False otherwise

update_meta(object_count, bytes_used, meta_timestamp=None)

Set the object stats metadata to the given values and update the meta_timestamp to the current time.

Parameters
  • object_count – should be an integer

  • bytes_used – should be an integer

  • meta_timestamp – timestamp for metadata; if not given the current time will be set.

Raises

ValueError – if object_count or bytes_used cannot be cast to an int, or if meta_timestamp is neither None nor can be cast to a Timestamp.

update_state(state, state_timestamp=None)

Set state to the given value and optionally update the state_timestamp to the given time.

Parameters
  • state – new state, should be an integer

  • state_timestamp – timestamp for state; if not given the state_timestamp will not be changed.

Returns

True if the state or state_timestamp was changed, False otherwise

update_tombstones(tombstones, meta_timestamp=None)

Set the tombstones metadata to the given values and update the meta_timestamp to the current time.

Parameters
  • tombstones – should be an integer

  • meta_timestamp – timestamp for metadata; if not given the current time will be set.

Raises

ValueError – if tombstones cannot be cast to an int, or if meta_timestamp is neither None nor can be cast to a Timestamp.

class swift.common.utils.ShardRangeList(initlist=None)

Bases: collections.UserList

This class provides some convenience functions for working with lists of ShardRange.

This class does not enforce ordering or continuity of the list items: callers should ensure that items are added in order as appropriate.

property bytes_used

Returns the total number of bytes in all items in the list.

Returns

total bytes used

filter(includes=None, marker=None, end_marker=None)

Filter the list for those shard ranges whose namespace includes the includes name or any part of the namespace between marker and end_marker. If none of includes, marker or end_marker are specified then all shard ranges will be returned.

Parameters
  • includes – a string; if not empty then only the shard range, if any, whose namespace includes this string will be returned, and marker and end_marker will be ignored.

  • marker – if specified then only shard ranges whose upper bound is greater than this value will be returned.

  • end_marker – if specified then only shard ranges whose lower bound is less than this value will be returned.

Returns

A new instance of ShardRangeList containing the filtered shard ranges.

find_lower(condition)

Finds the first shard range satisfies the given condition and returns its lower bound.

Parameters

condition – A function that must accept a single argument of type ShardRange and return True if the shard range satisfies the condition or False otherwise.

Returns

The lower bound of the first shard range to satisfy the condition, or the upper value of this list if no such shard range is found.

includes(other)

Check if another ShardRange namespace is enclosed between the list’s lower and upper properties. Note: the list’s lower and upper properties will only equal the outermost bounds of all items in the list if the list has previously been sorted.

Note: the list does not need to contain an item matching other for this method to return True, although if the list has been sorted and does contain an item matching other then the method will return True.

Parameters

other – an instance of ShardRange

Returns

True if other’s namespace is enclosed, False otherwise.

property lower

Returns the lower bound of the first item in the list. Note: this will only be equal to the lowest bound of all items in the list if the list contents has been sorted.

Returns

lower bound of first item in the list, or ShardRange.MIN if the list is empty.

property object_count

Returns the total number of objects of all items in the list.

Returns

total object count

property row_count

Returns the total number of rows of all items in the list.

Returns

total row count

property upper

Returns the upper bound of the first item in the list. Note: this will only be equal to the uppermost bound of all items in the list if the list has previously been sorted.

Returns

upper bound of first item in the list, or ShardRange.MIN if the list is empty.

class swift.common.utils.ShardRangeOuterBound

Bases: object

A custom singleton type to be subclassed for the outer bounds of ShardRanges.

class swift.common.utils.Spliterator(source_iterable)

Bases: object

Takes an iterator yielding sliceable things (e.g. strings or lists) and yields subiterators, each yielding up to the requested number of items from the source.

>>> si = Spliterator(["abcde", "fg", "hijkl"])
>>> ''.join(si.take(4))
"abcd"
>>> ''.join(si.take(3))
"efg"
>>> ''.join(si.take(1))
"h"
>>> ''.join(si.take(3))
"ijk"
>>> ''.join(si.take(3))
"l"  # shorter than requested; this can happen with the last iterator
class swift.common.utils.StrAnonymizer(data, method, salt)

Bases: str

Class that permits to get a string anonymized or simply quoted.

class swift.common.utils.StrFormatTime(ts)

Bases: object

Class that permits to get formats or parts of a time.

class swift.common.utils.StreamingPile(size)

Bases: swift.common.utils.GreenAsyncPile

Runs jobs in a pool of green threads, spawning more jobs as results are retrieved and worker threads become available.

When used as a context manager, has the same worker-killing properties as ContextPool.

asyncstarmap(func, args_iter)

This is the same as itertools.starmap(), except that func is executed in a separate green thread for each item, and results won’t necessarily have the same order as inputs.

class swift.common.utils.SwiftLogFormatter(fmt=None, datefmt=None, max_line_length=0)

Bases: logging.Formatter

Custom logging.Formatter will append txn_id to a log message if the record has one and the message does not. Optionally it can shorten overly long log lines.

format(record)

Format the specified record as text.

The record’s attribute dictionary is used as the operand to a string formatting operation which yields the returned string. Before formatting the dictionary, a couple of preparatory steps are carried out. The message attribute of the record is computed using LogRecord.getMessage(). If the formatting string uses the time (as determined by a call to usesTime(), formatTime() is called to format the event time. If there is exception information, it is formatted using formatException() and appended to the message.

class swift.common.utils.SwiftLoggerAdapter(logger, extra=None)

Bases: logging.LoggerAdapter

A logging.LoggerAdapter subclass that also passes through StatsD method calls.

Like logging.LoggerAdapter, you have to subclass this and override the process() method to accomplish anything useful.

exception(msg, *a, **kw)

Delegate an exception call to the underlying logger.

class swift.common.utils.ThreadSafeSysLogHandler(address=('localhost', 514), facility=1, socktype=None)

Bases: logging.handlers.SysLogHandler

createLock()

Acquire a thread lock for serializing access to the underlying I/O.

class swift.common.utils.Timestamp(timestamp, offset=0, delta=0, check_bounds=True)

Bases: object

Internal Representation of Swift Time.

The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like “1402464677.04188”

To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like “1402464677.04188_0000000000000000” - the portion after the underscore is the offset and is a formatted hexadecimal integer.

The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset.

The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it’s considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it’s offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset.

ceil()

Return the ‘normal’ part of the timestamp rounded up to the nearest integer number of seconds.

This value should be used whenever the second-precision Last-Modified time of a resource is required.

Returns

a float value with second precision.

classmethod from_isoformat(date_string)

Parse an isoformat string representation of time to a Timestamp object.

Parameters

date_string – a string formatted as per an Timestamp.isoformat property.

Returns

an instance of this class.

property isoformat

Get an isoformat string representation of the ‘normal’ part of the Timestamp with microsecond precision and no trailing timezone, for example:

1970-01-01T00:00:00.000000
Returns

an isoformat string

class swift.common.utils.Watchdog

Bases: object

Implements a watchdog to efficiently manage concurrent timeouts.

Compared to eventlet.timeouts.Timeout, it reduces the number of context switching in eventlet by avoiding to schedule actions (throw an Exception), then unschedule them if the timeouts are cancelled.

  1. at T+0, request timeout(10)

    => wathdog greenlet sleeps 10 seconds

  2. at T+1, request timeout(15)
    => the timeout will expire after the current, no need to wake up the

    watchdog greenlet

  3. at T+2, request timeout(5)
    => the timeout will expire before the first timeout, wake up the

    watchdog greenlet to calculate a new sleep period

  4. at T+7, the 3rd timeout expires
    => the exception is raised, then the greenlet watchdog sleep(3) to

    wake up for the 1st timeout expiration

spawn()

Start the watchdog greenthread.

start(timeout, exc, timeout_at=None)

Schedule a timeout action

Parameters
  • timeout – duration before the timeout expires

  • exc – exception to throw when the timeout expire, must inherit from eventlet.timeouts.Timeout

  • timeout_at – allow to force the expiration timestamp

Returns

id of the scheduled timeout, needed to cancel it

stop(key)

Cancel a scheduled timeout

Parameters

key – timeout id, as returned by start()

class swift.common.utils.WatchdogTimeout(watchdog, timeout, exc, timeout_at=None)

Bases: object

Context manager to schedule a timeout in a Watchdog instance

swift.common.utils.affinity_key_function(affinity_str)

Turns an affinity config value into a function suitable for passing to sort(). After doing so, the array will be sorted with respect to the given ordering.

For example, if affinity_str is “r1=1, r2z7=2, r2z8=2”, then the array will be sorted with all nodes from region 1 (r1=1) first, then all the nodes from region 2 zones 7 and 8 (r2z7=2 and r2z8=2), then everything else.

Note that the order of the pieces of affinity_str is irrelevant; the priority values are what comes after the equals sign.

If affinity_str is empty or all whitespace, then the resulting function will not alter the ordering of the nodes.

Parameters

affinity_str – affinity config value, e.g. “r1z2=3” or “r1=1, r2z1=2, r2z2=2”

Returns

single-argument function

Raises

ValueError – if argument invalid

swift.common.utils.affinity_locality_predicate(write_affinity_str)

Turns a write-affinity config value into a predicate function for nodes. The returned value will be a 1-arg function that takes a node dictionary and returns a true value if it is “local” and a false value otherwise. The definition of “local” comes from the affinity_str argument passed in here.

For example, if affinity_str is “r1, r2z2”, then only nodes where region=1 or where (region=2 and zone=2) are considered local.

If affinity_str is empty or all whitespace, then the resulting function will consider everything local

Parameters

write_affinity_str – affinity config value, e.g. “r1z2” or “r1, r2z1, r2z2”

Returns

single-argument function, or None if affinity_str is empty

Raises

ValueError – if argument invalid

swift.common.utils.audit_location_generator(devices, datadir, suffix='', mount_check=True, logger=None, devices_filter=None, partitions_filter=None, suffixes_filter=None, hashes_filter=None, hook_pre_device=None, hook_post_device=None, hook_pre_partition=None, hook_post_partition=None, hook_pre_suffix=None, hook_post_suffix=None, hook_pre_hash=None, hook_post_hash=None, error_counter=None, yield_hash_dirs=False)

Given a devices path and a data directory, yield (path, device, partition) for all files in that directory

(devices|partitions|suffixes|hashes)_filter are meant to modify the list of elements that will be iterated. eg: they can be used to exclude some elements based on a custom condition defined by the caller.

hook_pre_(device|partition|suffix|hash) are called before yielding the element, hook_pos_(device|partition|suffix|hash) are called after the element was yielded. They are meant to do some pre/post processing. eg: saving a progress status.

Parameters
  • devices – parent directory of the devices to be audited

  • datadir – a directory located under self.devices. This should be one of the DATADIR constants defined in the account, container, and object servers.

  • suffix – path name suffix required for all names returned (ignored if yield_hash_dirs is True)

  • mount_check – Flag to check if a mount check should be performed on devices

  • logger – a logger object

  • devices_filter – a callable taking (devices, [list of devices]) as parameters and returning a [list of devices]

  • partitions_filter – a callable taking (datadir_path, [list of parts]) as parameters and returning a [list of parts]

  • suffixes_filter – a callable taking (part_path, [list of suffixes]) as parameters and returning a [list of suffixes]

  • hashes_filter – a callable taking (suff_path, [list of hashes]) as parameters and returning a [list of hashes]

  • hook_pre_device – a callable taking device_path as parameter

  • hook_post_device – a callable taking device_path as parameter

  • hook_pre_partition – a callable taking part_path as parameter

  • hook_post_partition – a callable taking part_path as parameter

  • hook_pre_suffix – a callable taking suff_path as parameter

  • hook_post_suffix – a callable taking suff_path as parameter

  • hook_pre_hash – a callable taking hash_path as parameter

  • hook_post_hash – a callable taking hash_path as parameter

  • error_counter – a dictionary used to accumulate error counts; may add keys ‘unmounted’ and ‘unlistable_partitions’

  • yield_hash_dirs – if True, yield hash dirs instead of individual files

swift.common.utils.backward(f, blocksize=4096)

A generator returning lines from a file starting with the last line, then the second last line, etc. i.e., it reads lines backwards. Stops when the first line (if any) is read. This is useful when searching for recent activity in very large files.

Parameters
  • f – file object to read

  • blocksize – no of characters to go backwards at each block

swift.common.utils.cache_from_env(env, allow_none=False)

Get memcache connection pool from the environment (which had been previously set by the memcache middleware

Parameters

env – wsgi environment dict

Returns

swift.common.memcached.MemcacheRing from environment

swift.common.utils.capture_stdio(logger, **kwargs)

Log unhandled exceptions, close stdio, capture stdout and stderr.

param logger: Logger object to use

swift.common.utils.closing_if_possible(maybe_closable)

Like contextlib.closing(), but doesn’t crash if the object lacks a close() method.

PEP 333 (WSGI) says: “If the iterable returned by the application has a close() method, the server or gateway must call that method upon completion of the current request[.]” This function makes that easier.

swift.common.utils.compute_eta(start_time, current_value, final_value)

Compute an ETA. Now only if we could also have a progress bar…

Parameters
  • start_time – Unix timestamp when the operation began

  • current_value – Current value

  • final_value – Final value

Returns

ETA as a tuple of (length of time, unit of time) where unit of time is one of (‘h’, ‘m’, ‘s’)

swift.common.utils.config_auto_int_value(value, default)

Returns default if value is None or ‘auto’. Returns value as an int or raises ValueError otherwise.

swift.common.utils.config_fallocate_value(reserve_value)

Returns fallocate reserve_value as an int or float. Returns is_percent as a boolean. Returns a ValueError on invalid fallocate value.

swift.common.utils.config_positive_int_value(value)

Returns positive int value if it can be cast by int() and it’s an integer > 0. (not including zero) Raises ValueError otherwise.

swift.common.utils.config_read_prefixed_options(conf, prefix_name, defaults)

Read prefixed options from configuration

Parameters
  • conf – the configuration

  • prefix_name – the prefix (including, if needed, an underscore)

  • defaults – a dict of default values. The dict supplies the option name and type (string or comma separated string)

Returns

a dict containing the options

swift.common.utils.config_read_reseller_options(conf, defaults)

Read reseller_prefix option and associated options from configuration

Reads the reseller_prefix option, then reads options that may be associated with a specific reseller prefix. Reads options such that an option without a prefix applies to all reseller prefixes unless an option has an explicit prefix.

Parameters
  • conf – the configuration

  • defaults – a dict of default values. The key is the option name. The value is either an array of strings or a string

Returns

tuple of an array of reseller prefixes and a dict of option values

swift.common.utils.config_true_value(value)

Returns True if the value is either True or a string in TRUE_VALUES. Returns False otherwise.

swift.common.utils.csv_append(csv_string, item)

Appends an item to a comma-separated string.

If the comma-separated string is empty/None, just returns item.

swift.common.utils.decode_timestamps(encoded, explicit=False)

Parses a string of the form generated by encode_timestamps and returns a tuple of the three component timestamps. If explicit is False, component timestamps that are not explicitly encoded will be assumed to have zero delta from the previous component and therefore take the value of the previous component. If explicit is True, component timestamps that are not explicitly encoded will be returned with value None.

swift.common.utils.distribute_evenly(items, num_buckets)

Distribute items as evenly as possible into N buckets.

swift.common.utils.document_iters_to_http_response_body(ranges_iter, boundary, multipart, logger)

Takes an iterator of range iters and turns it into an appropriate HTTP response body, whether that’s multipart/byteranges or not.

This is almost, but not quite, the inverse of request_helpers.http_response_to_document_iters(). This function only yields chunks of the body, not any headers.

Parameters
  • ranges_iter

    an iterator of dictionaries, one per range. Each dictionary must contain at least the following key: “part_iter”: iterator yielding the bytes in the range

    Additionally, if multipart is True, then the following other keys are required:

    ”start_byte”: index of the first byte in the range “end_byte”: index of the last byte in the range “content_type”: value for the range’s Content-Type header

    Finally, there is one optional key that is used in the

    multipart/byteranges case:

    ”entity_length”: length of the requested entity (not necessarily

    equal to the response length). If omitted, “*” will be used.

    Each part_iter will be exhausted prior to calling next(ranges_iter).

  • boundary – MIME boundary to use, sans dashes (e.g. “boundary”, not “–boundary”).

  • multipart – True if the response should be multipart/byteranges, False otherwise. This should be True if and only if you have 2 or more ranges.

  • logger – a logger

swift.common.utils.document_iters_to_multipart_byteranges(ranges_iter, boundary)

Takes an iterator of range iters and yields a multipart/byteranges MIME document suitable for sending as the body of a multi-range 206 response.

See document_iters_to_http_response_body for parameter descriptions.

swift.common.utils.drain_and_close(response_or_app_iter)

Drain and close a swob or WSGI response.

This ensures we don’t log a 499 in the proxy just because we realized we don’t care about the body of an error.

swift.common.utils.drop_buffer_cache(fd, offset, length)

Drop ‘buffer’ cache for the given range of the given file.

Parameters
  • fd – file descriptor

  • offset – start offset

  • length – length

swift.common.utils.drop_privileges(user)

Sets the userid/groupid of the current process, get session leader, etc.

Parameters

user – User name to change privileges to

swift.common.utils.dump_recon_cache(cache_dict, cache_file, logger, lock_timeout=2, set_owner=None)

Update recon cache values

Parameters
  • cache_dict – Dictionary of cache key/value pairs to write out

  • cache_file – cache file to update

  • logger – the logger to use to log an encountered error

  • lock_timeout – timeout (in seconds)

  • set_owner – Set owner of recon cache file

swift.common.utils.encode_timestamps(t1, t2=None, t3=None, explicit=False)

Encode up to three timestamps into a string. Unlike a Timestamp object, the encoded string does NOT used fixed width fields and consequently no relative chronology of the timestamps can be inferred from lexicographic sorting of encoded timestamp strings.

The format of the encoded string is:

<t1>[<+/-><t2 - t1>[<+/-><t3 - t2>]]

i.e. if t1 = t2 = t3 then just the string representation of t1 is returned, otherwise the time offsets for t2 and t3 are appended. If explicit is True then the offsets for t2 and t3 are always appended even if zero.

Note: any offset value in t1 will be preserved, but offsets on t2 and t3 are not preserved. In the anticipated use cases for this method (and the inverse decode_timestamps method) the timestamps passed as t2 and t3 are not expected to have offsets as they will be timestamps associated with a POST request. In the case where the encoding is used in a container objects table row, t1 could be the PUT or DELETE time but t2 and t3 represent the content type and metadata times (if different from the data file) i.e. correspond to POST timestamps. In the case where the encoded form is used in a .meta file name, t1 and t2 both correspond to POST timestamps.

swift.common.utils.eventlet_monkey_patch()

Install the appropriate Eventlet monkey patches.

swift.common.utils.expand_ipv6(address)

Expand ipv6 address. :param address: a string indicating valid ipv6 address :returns: a string indicating fully expanded ipv6 address

swift.common.utils.extract_swift_bytes(content_type)
Parse a content-type and return a tuple containing:
  • the content_type string minus any swift_bytes param,

  • the swift_bytes value or None if the param was not found

Parameters

content_type – a content-type string

Returns

a tuple of (content-type, swift_bytes or None)

swift.common.utils.fallocate(fd, size, offset=0)

Pre-allocate disk space for a file.

This function can be disabled by calling disable_fallocate(). If no suitable C function is available in libc, this function is a no-op.

Parameters
  • fd – file descriptor

  • size – size to allocate (in bytes)

swift.common.utils.fdatasync(fd)

Sync modified file data to disk.

Parameters

fd – file descriptor

swift.common.utils.filter_shard_ranges(shard_ranges, includes, marker, end_marker)

Filter the given shard ranges to those whose namespace includes the includes name or any part of the namespace between marker and end_marker. If none of includes, marker or end_marker are specified then all shard ranges will be returned.

Parameters
  • shard_ranges – A list of ShardRange.

  • includes – a string; if not empty then only the shard range, if any, whose namespace includes this string will be returned, and marker and end_marker will be ignored.

  • marker – if specified then only shard ranges whose upper bound is greater than this value will be returned.

  • end_marker – if specified then only shard ranges whose lower bound is less than this value will be returned.

Returns

A filtered list of ShardRange.

swift.common.utils.find_shard_range(item, ranges)

Find a ShardRange in given list of shard_ranges whose namespace contains item.

Parameters
  • item – The item for a which a ShardRange is to be found.

  • ranges – a sorted list of ShardRanges.

Returns

the ShardRange whose namespace contains item, or None if no suitable range is found.

swift.common.utils.fs_has_free_space(fs_path, space_needed, is_percent)

Check to see whether or not a filesystem has the given amount of space free. Unlike fallocate(), this does not reserve any space.

Parameters
  • fs_path – path to a file or directory on the filesystem; typically the path to the filesystem’s mount point

  • space_needed – minimum bytes or percentage of free space

  • is_percent – if True, then space_needed is treated as a percentage of the filesystem’s capacity; if False, space_needed is a number of free bytes.

Returns

True if the filesystem has at least that much free space, False otherwise

Raises

OSError – if fs_path does not exist

swift.common.utils.fsync(fd)

Sync modified file data and metadata to disk.

Parameters

fd – file descriptor

swift.common.utils.fsync_dir(dirpath)

Sync directory entries to disk.

Parameters

dirpath – Path to the directory to be synced.

swift.common.utils.get_db_files(db_path)

Given the path to a db file, return a sorted list of all valid db files that actually exist in that path’s dir. A valid db filename has the form:

<hash>[_<epoch>].db

where <hash> matches the <hash> part of the given db_path as would be parsed by parse_db_filename().

Parameters

db_path – Path to a db file that does not necessarily exist.

Returns

List of valid db files that do exist in the dir of the db_path. This list may be empty.

swift.common.utils.get_expirer_container(x_delete_at, expirer_divisor, acc, cont, obj)

Returns an expiring object container name for given X-Delete-At and (native string) a/c/o.

swift.common.utils.get_hub()

Checks whether poll is available and falls back on select if it isn’t.

Note about epoll:

Review: https://review.opendev.org/#/c/18806/

There was a problem where once out of every 30 quadrillion connections, a coroutine wouldn’t wake up when the client closed its end. Epoll was not reporting the event or it was getting swallowed somewhere. Then when that file descriptor was re-used, eventlet would freak right out because it still thought it was waiting for activity from it in some other coro.

Another note about epoll: it’s hard to use when forking. epoll works like so:

  • create an epoll instance: efd = epoll_create(...)

  • register file descriptors of interest with epoll_ctl(efd, EPOLL_CTL_ADD, fd, ...)

  • wait for events with epoll_wait(efd, ...)

If you fork, you and all your child processes end up using the same epoll instance, and everyone becomes confused. It is possible to use epoll and fork and still have a correct program as long as you do the right things, but eventlet doesn’t do those things. Really, it can’t even try to do those things since it doesn’t get notified of forks.

In contrast, both poll() and select() specify the set of interesting file descriptors with each call, so there’s no problem with forking.

As eventlet monkey patching is now done before call get_hub() in wsgi.py if we use ‘import select’ we get the eventlet version, but since version 0.20.0 eventlet removed select.poll() function in patched select (see: http://eventlet.net/doc/changelog.html and https://github.com/eventlet/eventlet/commit/614a20462).

We use eventlet.patcher.original function to get python select module to test if poll() is available on platform.

swift.common.utils.get_log_line(req, res, trans_time, additional_info, fmt, anonymization_method, anonymization_salt)

Make a line for logging that matches the documented log line format for backend servers.

Parameters
  • req – the request.

  • res – the response.

  • trans_time – the time the request took to complete, a float.

  • additional_info – a string to log at the end of the line

Returns

a properly formatted line for logging.

swift.common.utils.get_logger(conf, name=None, log_to_console=False, log_route=None, fmt='%(server)s: %(message)s', statsd_tail_prefix=None)

Get the current system logger using config settings.

Log config and defaults:

log_facility = LOG_LOCAL0
log_level = INFO
log_name = swift
log_max_line_length = 0
log_udp_host = (disabled)
log_udp_port = logging.handlers.SYSLOG_UDP_PORT
log_address = /dev/log
log_statsd_host = (disabled)
log_statsd_port = 8125
log_statsd_default_sample_rate = 1.0
log_statsd_sample_rate_factor = 1.0
log_statsd_metric_prefix = (empty-string)
Parameters
  • conf – Configuration dict to read settings from

  • name – This value is used to populate the server field in the log format, as the prefix for statsd messages, and as the default value for log_route; defaults to the log_name value in conf, if it exists, or to ‘swift’.

  • log_to_console – Add handler which writes to console on stderr

  • log_route – Route for the logging, not emitted to the log, just used to separate logging configurations; defaults to the value of name or whatever name defaults to. This value is used as the name attribute of the logging.LogAdapter that is returned.

  • fmt – Override log format

  • statsd_tail_prefix – tail prefix to pass to statsd client; if None then the tail prefix defaults to the value of name.

Returns

an instance of LogAdapter

swift.common.utils.get_md5_socket()

Get an MD5 socket file descriptor. One can MD5 data with it by writing it to the socket with os.write, then os.read the 16 bytes of the checksum out later.

NOTE: It is the caller’s responsibility to ensure that os.close() is called on the returned file descriptor. This is a bare file descriptor, not a Python object. It doesn’t close itself.

swift.common.utils.get_partition_for_hash(hex_hash, part_power)

Return partition number for given hex hash and partition power. :param hex_hash: A hash string :param part_power: partition power :returns: partition number

swift.common.utils.get_partition_from_path(devices, path)
Parameters
  • devices – directory where devices are mounted (e.g. /srv/node)

  • path – full path to a object file or hashdir

Returns

the (integer) partition from the path

swift.common.utils.get_policy_index(req_headers, res_headers)

Returns the appropriate index of the storage policy for the request from a proxy server

Parameters
  • req_headers – dict of the request headers.

  • res_headers – dict of the response headers.

Returns

string index of storage policy, or None

swift.common.utils.get_redirect_data(response)

Extract a redirect location from a response’s headers.

Parameters

response – a response

Returns

a tuple of (path, Timestamp) if a Location header is found, otherwise None

Raises

ValueError – if the Location header is found but a X-Backend-Redirect-Timestamp is not found, or if there is a problem with the format of etiher header

swift.common.utils.get_time_units(time_amount)

Get a nomralized length of time in the largest unit of time (hours, minutes, or seconds.)

Parameters

time_amount – length of time in seconds

Returns

A touple of (length of time, unit of time) where unit of time is one of (‘h’, ‘m’, ‘s’)

swift.common.utils.get_valid_utf8_str(str_or_unicode)

Get valid parts of utf-8 str from str, unicode and even invalid utf-8 str

Parameters

str_or_unicode – a string or an unicode which can be invalid utf-8

swift.common.utils.get_zero_indexed_base_string(base, index)

This allows the caller to make a list of things with indexes, where the first item (zero indexed) is just the bare base string, and subsequent indexes are appended ‘-1’, ‘-2’, etc.

e.g.:

'lock', None => 'lock'
'lock', 0    => 'lock'
'lock', 1    => 'lock-1'
'object', 2  => 'object-2'
Parameters
  • base – a string, the base string; when index is 0 (or None) this is the identity function.

  • index – a digit, typically an integer (or None); for values other than 0 or None this digit is appended to the base string separated by a hyphen.

swift.common.utils.hash_path(account, container=None, object=None, raw_digest=False)

Get the canonical hash for an account/container/object

Parameters
  • account – Account

  • container – Container

  • object – Object

  • raw_digest – If True, return the raw version rather than a hex digest

Returns

hash string

swift.common.utils.human_readable(value)

Returns the number in a human readable format; for example 1048576 = “1Mi”.

swift.common.utils.is_file_older(path, age)

Test if a file mtime is older than the given age, suppressing any OSErrors.

Parameters
  • path – first and only argument passed to os.stat

  • age – age in seconds

Returns

True if age is less than or equal to zero or if the file mtime is more than age in the past; False if age is greater than zero and the file mtime is less than or equal to age in the past or if there is an OSError while stat’ing the file.

swift.common.utils.is_valid_ip(ip)

Return True if the provided ip is a valid IP-address

swift.common.utils.is_valid_ipv4(ip)

Return True if the provided ip is a valid IPv4-address

swift.common.utils.is_valid_ipv6(ip)

Returns True if the provided ip is a valid IPv6-address

swift.common.utils.ismount(path)

Test whether a path is a mount point. This will catch any exceptions and translate them into a False return value Use ismount_raw to have the exceptions raised instead.

swift.common.utils.ismount_raw(path)

Test whether a path is a mount point. Whereas ismount will catch any exceptions and just return False, this raw version will not catch exceptions.

This is code hijacked from C Python 2.6.8, adapted to remove the extra lstat() system call.

swift.common.utils.item_from_env(env, item_name, allow_none=False)

Get a value from the wsgi environment

Parameters
  • env – wsgi environment dict

  • item_name – name of item to get

Returns

the value from the environment

swift.common.utils.iter_multipart_mime_documents(wsgi_input, boundary, read_chunk_size=4096)

Given a multi-part-mime-encoded input file object and boundary, yield file-like objects for each part. Note that this does not split each part into headers and body; the caller is responsible for doing that if necessary.

Parameters
  • wsgi_input – The file-like object to read from.

  • boundary – The mime boundary to separate new file-like objects on.

Returns

A generator of file-like objects for each part.

Raises

MimeInvalid – if the document is malformed

swift.common.utils.last_modified_date_to_timestamp(last_modified_date_str)

Convert a last modified date (like you’d get from a container listing, e.g. 2014-02-28T23:22:36.698390) to a float.

Creates a link to file descriptor at target_path specified. This method does not close the fd for you. Unlike rename, as linkat() cannot overwrite target_path if it exists, we unlink and try again.

Attempts to fix / hide race conditions like empty object directories being removed by backend processes during uploads, by retrying.

Parameters
  • fd – File descriptor to be linked

  • target_path – Path in filesystem where fd is to be linked

  • dirs_created – Number of newly created directories that needs to be fsync’d.

  • retries – number of retries to make

  • fsync – fsync on containing directory of target_path and also all the newly created directories.

swift.common.utils.list_from_csv(comma_separated_str)

Splits the str given and returns a properly stripped list of the comma separated values.

swift.common.utils.load_libc_function(func_name, log_error=True, fail_if_missing=False, errcheck=False)

Attempt to find the function in libc, otherwise return a no-op func.

Parameters
  • func_name – name of the function to pull from libc.

  • log_error – log an error when a function can’t be found

  • fail_if_missing – raise an exception when a function can’t be found. Default behavior is to return a no-op function.

  • errcheck – boolean, if true install a wrapper on the function to check for a return values of -1 and call ctype.get_errno and raise an OSError

swift.common.utils.load_recon_cache(cache_file)

Load a recon cache file. Treats missing file as empty.

swift.common.utils.lock_file(filename, timeout=None, append=False, unlink=True)

Context manager that acquires a lock on a file. This will block until the lock can be acquired, or the timeout time has expired (whichever occurs first).

Parameters
  • filename – file to be locked

  • timeout – timeout (in seconds). If None, defaults to DEFAULT_LOCK_TIMEOUT

  • append – True if file should be opened in append mode

  • unlink – True if the file should be unlinked at the end

swift.common.utils.lock_parent_directory(filename, timeout=None)

Context manager that acquires a lock on the parent directory of the given file path. This will block until the lock can be acquired, or the timeout time has expired (whichever occurs first).

Parameters
  • filename – file path of the parent directory to be locked

  • timeout – timeout (in seconds). If None, defaults to DEFAULT_LOCK_TIMEOUT

swift.common.utils.lock_path(directory, timeout=None, timeout_class=None, limit=1, name=None)

Context manager that acquires a lock on a directory. This will block until the lock can be acquired, or the timeout time has expired (whichever occurs first).

For locking exclusively, file or directory has to be opened in Write mode. Python doesn’t allow directories to be opened in Write Mode. So we workaround by locking a hidden file in the directory.

Parameters
  • directory – directory to be locked

  • timeout – timeout (in seconds). If None, defaults to DEFAULT_LOCK_TIMEOUT

  • timeout_class – The class of the exception to raise if the lock cannot be granted within the timeout. Will be constructed as timeout_class(timeout, lockpath). Default: LockTimeout

  • limit – The maximum number of locks that may be held concurrently on the same directory at the time this method is called. Note that this limit is only applied during the current call to this method and does not prevent subsequent calls giving a larger limit. Defaults to 1.

  • name – A string to distinguishes different type of locks in a directory

Raises
  • TypeError – if limit is not an int.

  • ValueError – if limit is less than 1.

swift.common.utils.make_db_file_path(db_path, epoch)

Given a path to a db file, return a modified path whose filename part has the given epoch.

A db filename takes the form <hash>[_<epoch>].db; this method replaces the <epoch> part of the given db_path with the given epoch value, or drops the epoch part if the given epoch is None.

Parameters
  • db_path – Path to a db file that does not necessarily exist.

  • epoch – A string (or None) that will be used as the epoch in the new path’s filename; non-None values will be normalized to the normal string representation of a Timestamp.

Returns

A modified path to a db file.

Raises

ValueError – if the epoch is not valid for constructing a Timestamp.

swift.common.utils.makedirs_count(path, count=0)

Same as os.makedirs() except that this method returns the number of new directories that had to be created.

Also, this does not raise an error if target directory already exists. This behaviour is similar to Python 3.x’s os.makedirs() called with exist_ok=True. Also similar to swift.common.utils.mkdirs()

https://hg.python.org/cpython/file/v3.4.2/Lib/os.py#l212

swift.common.utils.maybe_multipart_byteranges_to_document_iters(app_iter, content_type)

Takes an iterator that may or may not contain a multipart MIME document as well as content type and returns an iterator of body iterators.

Parameters
  • app_iter – iterator that may contain a multipart MIME document

  • content_type – content type of the app_iter, used to determine whether it conains a multipart document and, if so, what the boundary is between documents

swift.common.utils.md5(string=b'', usedforsecurity=True)

Return an md5 hashlib object using usedforsecurity parameter

For python distributions that support the usedforsecurity keyword parameter, this passes the parameter through as expected. See https://bugs.python.org/issue9216

swift.common.utils.md5_hash_for_file(fname)

Get the MD5 checksum of a file.

Parameters

fname – path to file

Returns

MD5 checksum, hex encoded

swift.common.utils.memcached_timing_stats(**dec_kwargs)

Returns a decorator that logs timing events or errors for public methods in MemcacheRing class, such as memcached set, get and etc.

swift.common.utils.mime_to_document_iters(input_file, boundary, read_chunk_size=4096)

Takes a file-like object containing a multipart MIME document and returns an iterator of (headers, body-file) tuples.

Parameters
  • input_file – file-like object with the MIME doc in it

  • boundary – MIME boundary, sans dashes (e.g. “divider”, not “–divider”)

  • read_chunk_size – size of strings read via input_file.read()

swift.common.utils.mkdirs(path)

Ensures the path is a directory or makes it if not. Errors if the path exists but is a file or on permissions failure.

Parameters

path – path to create

swift.common.utils.modify_priority(conf, logger)

Modify priority by nice and ionice.

swift.common.utils.multipart_byteranges_to_document_iters(input_file, boundary, read_chunk_size=4096)

Takes a file-like object containing a multipart/byteranges MIME document (see RFC 7233, Appendix A) and returns an iterator of (first-byte, last-byte, length, document-headers, body-file) 5-tuples.

Parameters
  • input_file – file-like object with the MIME doc in it

  • boundary – MIME boundary, sans dashes (e.g. “divider”, not “–divider”)

  • read_chunk_size – size of strings read via input_file.read()

swift.common.utils.node_to_string(node_dict, replication=False)

Get a string representation of a node’s location.

Parameters
  • node_dict – a dict describing a node

  • replication – if True then the replication ip address and port are used, otherwise the normal ip address and port are used.

Returns

a string of the form <ip address>:<port>/<device>

swift.common.utils.non_negative_float(value)

Check that the value casts to a float and is non-negative.

Parameters

value – value to check

Raises

ValueError – if the value cannot be cast to a float or is negative.

Returns

a float

swift.common.utils.non_negative_int(value)

Check that the value casts to an int and is a whole number.

Parameters

value – value to check

Raises

ValueError – if the value cannot be cast to an int or does not represent a whole number.

Returns

an int

swift.common.utils.normalize_delete_at_timestamp(timestamp, high_precision=False)

Format a timestamp (string or numeric) into a standardized xxxxxxxxxx (10) or xxxxxxxxxx.xxxxx (10.5) format.

Note that timestamps less than 0000000000 are raised to 0000000000 and values greater than November 20th, 2286 at 17:46:39 UTC will be capped at that date and time, resulting in no return value exceeding 9999999999.99999 (or 9999999999 if using low-precision).

This cap is because the expirer is already working through a sorted list of strings that were all a length of 10. Adding another digit would mess up the sort and cause the expirer to break from processing early. By 2286, this problem will need to be fixed, probably by creating an additional .expiring_objects account to work from with 11 (or more) digit container names.

Parameters

timestamp – unix timestamp

Returns

normalized timestamp as a string

swift.common.utils.normalize_timestamp(timestamp)

Format a timestamp (string or numeric) into a standardized xxxxxxxxxx.xxxxx (10.5) format.

Note that timestamps using values greater than or equal to November 20th, 2286 at 17:46 UTC will use 11 digits to represent the number of seconds.

Parameters

timestamp – unix timestamp

Returns

normalized timestamp as a string

swift.common.utils.override_bytes_from_content_type(listing_dict, logger=None)

Takes a dict from a container listing and overrides the content_type, bytes fields if swift_bytes is set.

swift.common.utils.pairs(item_list)

Returns an iterator of all pairs of elements from item_list.

Parameters

item_list – items (no duplicates allowed)

swift.common.utils.parse_content_disposition(header)

Given the value of a header like: Content-Disposition: form-data; name=”somefile”; filename=”test.html”

Return data like (“form-data”, {“name”: “somefile”, “filename”: “test.html”})

Parameters

header – Value of a header (the part after the ‘: ‘).

Returns

(value name, dict) of the attribute data parsed (see above).

swift.common.utils.parse_content_range(content_range)

Parse a content-range header into (first_byte, last_byte, total_size).

See RFC 7233 section 4.2 for details on the header format, but it’s basically “Content-Range: bytes ${start}-${end}/${total}”.

Parameters

content_range – Content-Range header value to parse, e.g. “bytes 100-1249/49004”

Returns

3-tuple (start, end, total)

Raises

ValueError – if malformed

swift.common.utils.parse_content_type(content_type)

Parse a content-type and its parameters into values. RFC 2616 sec 14.17 and 3.7 are pertinent.

Examples:

'text/plain; charset=UTF-8' -> ('text/plain', [('charset, 'UTF-8')])
'text/plain; charset=UTF-8; level=1' ->
    ('text/plain', [('charset, 'UTF-8'), ('level', '1')])
Parameters

content_type – content_type to parse

Returns

a tuple containing (content type, list of k, v parameter tuples)

swift.common.utils.parse_db_filename(filename)

Splits a db filename into three parts: the hash, the epoch, and the extension.

>>> parse_db_filename("ab2134.db")
('ab2134', None, '.db')
>>> parse_db_filename("ab2134_1234567890.12345.db")
('ab2134', '1234567890.12345', '.db')
Parameters

filename – A db file basename or path to a db file.

Returns

A tuple of (hash , epoch, extension). epoch may be None.

Raises

ValueError – if filename is not a path to a file.

swift.common.utils.parse_mime_headers(doc_file)

Takes a file-like object containing a MIME document and returns a HeaderKeyDict containing the headers. The body of the message is not consumed: the position in doc_file is left at the beginning of the body.

This function was inspired by the Python standard library’s http.client.parse_headers.

Parameters

doc_file – binary file-like object containing a MIME document

Returns

a swift.common.swob.HeaderKeyDict containing the headers

swift.common.utils.parse_options(parser=None, once=False, test_args=None)

Parse standard swift server/daemon options with optparse.OptionParser.

Parameters
  • parser – OptionParser to use. If not sent one will be created.

  • once – Boolean indicating the “once” option is available

  • test_args – Override sys.argv; used in testing

Returns

Tuple of (config, options); config is an absolute path to the config file, options is the parser options as a dictionary.

Raises

SystemExit – First arg (CONFIG) is required, file must exist

swift.common.utils.parse_override_options(**kwargs)

Figure out which policies, devices, and partitions we should operate on, based on kwargs.

If ‘override_policies’ is already present in kwargs, then return that value. This happens when using multiple worker processes; the parent process supplies override_policies=X to each child process.

Otherwise, in run-once mode, look at the ‘policies’ keyword argument. This is the value of the “–policies” command-line option. In run-forever mode or if no –policies option was provided, an empty list will be returned.

The procedures for devices and partitions are similar.

Returns

a named tuple with fields “devices”, “partitions”, and “policies”.

swift.common.utils.parse_prefixed_conf(conf_file, prefix)

Search the config file for any common-prefix sections and load those sections to a dict mapping the after-prefix reference to options.

Parameters
  • conf_file – the file name of the config to parse

  • prefix – the common prefix of the sections

Returns

a dict mapping policy reference -> dict of policy options

Raises

ValueError – if a policy config section has an invalid name

swift.common.utils.parse_socket_string(socket_string, default_port)

Given a string representing a socket, returns a tuple of (host, port). Valid strings are DNS names, IPv4 addresses, or IPv6 addresses, with an optional port. If an IPv6 address is specified it must be enclosed in [], like [::1] or [::1]:11211. This follows the accepted prescription for IPv6 host literals.

Examples:

server.org
server.org:1337
127.0.0.1:1337
[::1]:1337
[::1]
swift.common.utils.private(func)

Decorator to declare which methods are privately accessible as HTTP requests with an X-Backend-Allow-Private-Methods: True override

Parameters

func – function to make private

swift.common.utils.public(func)

Decorator to declare which methods are publicly accessible as HTTP requests

Parameters

func – function to make public

swift.common.utils.punch_hole(fd, offset, length)

De-allocate disk space in the middle of a file.

Parameters
  • fd – file descriptor

  • offset – index of first byte to de-allocate

  • length – number of bytes to de-allocate

swift.common.utils.put_recon_cache_entry(cache_entry, key, item)

Update a recon cache entry item.

If item is an empty dict then any existing key in cache_entry will be deleted. Similarly if item is a dict and any of its values are empty dicts then the corrsponsing key will be deleted from the nested dict in cache_entry.

We use nested recon cache entries when the object auditor runs in parallel or else in ‘once’ mode with a specified subset of devices.

Parameters
  • cache_entry – a dict of existing cache entries

  • key – key for item to update

  • item – value for item to update

swift.common.utils.quorum_size(n)

quorum size as it applies to services that use ‘replication’ for data integrity (Account/Container services). Object quorum_size is defined on a storage policy basis.

Number of successful backend requests needed for the proxy to consider the client request successful.

swift.common.utils.quote(value, safe='/')

Patched version of urllib.quote that encodes utf-8 strings before quoting

swift.common.utils.random() x in the interval [0, 1).
swift.common.utils.ratelimit_sleep(running_time, max_rate, incr_by=1, rate_buffer=5)

Will eventlet.sleep() for the appropriate time so that the max_rate is never exceeded. If max_rate is 0, will not ratelimit. The maximum recommended rate should not exceed (1000 * incr_by) a second as eventlet.sleep() does involve some overhead. Returns running_time that should be used for subsequent calls.

Parameters
  • running_time – the running time in milliseconds of the next allowable request. Best to start at zero.

  • max_rate – The maximum rate per second allowed for the process.

  • incr_by – How much to increment the counter. Useful if you want to ratelimit 1024 bytes/sec and have differing sizes of requests. Must be > 0 to engage rate-limiting behavior.

  • rate_buffer – Number of seconds the rate counter can drop and be allowed to catch up (at a faster than listed rate). A larger number will result in larger spikes in rate but better average accuracy. Must be > 0 to engage rate-limiting behavior.

Returns

The absolute time for the next interval in milliseconds; note that time could have passed well beyond that point, but the next call will catch that and skip the sleep.

swift.common.utils.readconf(conf_path, section_name=None, log_name=None, defaults=None, raw=False)

Read config file(s) and return config items as a dict

Parameters
  • conf_path – path to config file/directory, or a file-like object (hasattr readline)

  • section_name – config section to read (will return all sections if not defined)

  • log_name – name to be used with logging (will use section_name if not defined)

  • defaults – dict of default values to pre-populate the config with

Returns

dict of config items

Raises
  • ValueError – if section_name does not exist

  • IOError – if reading the file failed

swift.common.utils.reiterate(iterable)

Consume the first truthy item from an iterator, then re-chain it to the rest of the iterator. This is useful when you want to make sure the prologue to downstream generators have been executed before continuing. :param iterable: an iterable object

swift.common.utils.remove_directory(path)

Wrapper for os.rmdir, ENOENT and ENOTEMPTY are ignored

Parameters

path – first and only argument passed to os.rmdir

swift.common.utils.remove_file(path)

Quiet wrapper for os.unlink, OSErrors are suppressed

Parameters

path – first and only argument passed to os.unlink

swift.common.utils.renamer(old, new, fsync=True)

Attempt to fix / hide race conditions like empty object directories being removed by backend processes during uploads, by retrying.

The containing directory of ‘new’ and of all newly created directories are fsync’d by default. This _will_ come at a performance penalty. In cases where these additional fsyncs are not necessary, it is expected that the caller of renamer() turn it off explicitly.

Parameters
  • old – old path to be renamed

  • new – new path to be renamed to

  • fsync – fsync on containing directory of new and also all the newly created directories.

swift.common.utils.replace_partition_in_path(devices, path, part_power)

Takes a path and a partition power and returns the same path, but with the correct partition number. Most useful when increasing the partition power.

Parameters
  • devices – directory where devices are mounted (e.g. /srv/node)

  • path – full path to a object file or hashdir

  • part_power – partition power to compute correct partition number

Returns

Path with re-computed partition power

swift.common.utils.replication(func)

Decorator to declare which methods are accessible for different type of servers:

  • If option replication_server is None then this decorator doesn’t matter.

  • If option replication_server is True then ONLY decorated with this decorator methods will be started.

  • If option replication_server is False then decorated with this decorator methods will NOT be started.

Parameters

func – function to mark accessible for replication

swift.common.utils.round_robin_iter(its)

Takes a list of iterators, yield an element from each in a round-robin fashion until all of them are exhausted. :param its: list of iterators

swift.common.utils.rsync_ip(ip)

Transform ip string to an rsync-compatible form

Will return ipv4 addresses unchanged, but will nest ipv6 addresses inside square brackets.

Parameters

ip – an ip string (ipv4 or ipv6)

Returns

a string ip address

swift.common.utils.rsync_module_interpolation(template, device)

Interpolate devices variables inside a rsync module template

Parameters
  • template – rsync module template as a string

  • device – a device from a ring

Returns

a string with all variables replaced by device attributes

swift.common.utils.search_tree(root, glob_match, ext='', exts=None, dir_ext=None)

Look in root, for any files/dirs matching glob, recursively traversing any found directories looking for files ending with ext

Parameters
  • root – start of search path

  • glob_match – glob to match in root, matching dirs are traversed with os.walk

  • ext – only files that end in ext will be returned

  • exts – a list of file extensions; only files that end in one of these extensions will be returned; if set this list overrides any extension specified using the ‘ext’ param.

  • dir_ext – if present directories that end with dir_ext will not be traversed and instead will be returned as a matched path

Returns

list of full paths to matching files, sorted

swift.common.utils.select_ip_port(node_dict, use_replication=False)

Get the ip address and port that should be used for the given node_dict.

If use_replication is True then the replication ip address and port are returned.

If use_replication is False (the default) and the node dict has an item with key use_replication then that item’s value will determine if the replication ip address and port are returned.

If neither use_replication nor node_dict['use_replication'] indicate otherwise then the normal ip address and port are returned.

Parameters
  • node_dict – a dict describing a node

  • use_replication – if True then the replication ip address and port are returned.

Returns

a tuple of (ip address, port)

swift.common.utils.set_swift_dir(swift_dir)

Sets the directory from which swift config files will be read. If the given directory differs from that already set then the swift.conf file in the new directory will be validated and storage policies will be reloaded from the new swift.conf file.

Parameters

swift_dir – non-default directory to read swift.conf from

class swift.common.utils.sockaddr_alg

Bases: _ctypes.Structure

swift.common.utils.split_path(path, minsegs=1, maxsegs=None, rest_with_last=False)

Validate and split the given HTTP request path.

Examples:

['a'] = split_path('/a')
['a', None] = split_path('/a', 1, 2)
['a', 'c'] = split_path('/a/c', 1, 2)
['a', 'c', 'o/r'] = split_path('/a/c/o/r', 1, 3, True)
Parameters
  • path – HTTP Request path to be split

  • minsegs – Minimum number of segments to be extracted

  • maxsegs – Maximum number of segments to be extracted

  • rest_with_last – If True, trailing data will be returned as part of last segment. If False, and there is trailing data, raises ValueError.

Returns

list of segments with a length of maxsegs (non-existent segments will return as None)

Raises

ValueError – if given an invalid path

swift.common.utils.storage_directory(datadir, partition, name_hash)

Get the storage directory

Parameters
  • datadir – Base data directory

  • partition – Partition

  • name_hash – Account, container or object name hash

Returns

Storage directory

swift.common.utils.streq_const_time(s1, s2)

Constant-time string comparison.

Params s1

the first string

Params s2

the second string

Returns

True if the strings are equal.

This function takes two strings and compares them. It is intended to be used when doing a comparison for authentication purposes to help guard against timing attacks.

swift.common.utils.strict_b64decode(value, allow_line_breaks=False)

Validate and decode Base64-encoded data.

The stdlib base64 module silently discards bad characters, but we often want to treat them as an error.

Parameters
  • value – some base64-encoded data

  • allow_line_breaks – if True, ignore carriage returns and newlines

Returns

the decoded data

Raises

ValueError – if value is not a string, contains invalid characters, or has insufficient padding

swift.common.utils.systemd_notify(logger=None)

Notify the service manager that started this process, if it is systemd-compatible, that this process correctly started. To do so, it communicates through a Unix socket stored in environment variable NOTIFY_SOCKET. More information can be found in systemd documentation: https://www.freedesktop.org/software/systemd/man/sd_notify.html

Parameters

logger – a logger object

swift.common.utils.timing_stats(**dec_kwargs)

Returns a decorator that logs timing events or errors for public methods in swift’s wsgi server controllers, based on response code.

Remove any file in a given path that was last modified before mtime.

Parameters
  • path – path to remove file from

  • mtime – timestamp of oldest file to keep

Remove any files from the given list that were last modified before mtime.

Parameters
  • filepaths – a list of strings, the full paths of files to check

  • mtime – timestamp of oldest file to keep

swift.common.utils.validate_device_partition(device, partition)

Validate that a device and a partition are valid and won’t lead to directory traversal when used.

Parameters
  • device – device to validate

  • partition – partition to validate

Raises

ValueError – if given an invalid device or partition

swift.common.utils.validate_sync_to(value, allowed_sync_hosts, realms_conf)

Validates an X-Container-Sync-To header value, returning the validated endpoint, realm, and realm_key, or an error string.

Parameters
  • value – The X-Container-Sync-To header value to validate.

  • allowed_sync_hosts – A list of allowed hosts in endpoints, if realms_conf does not apply.

  • realms_conf – An instance of swift.common.container_sync_realms.ContainerSyncRealms to validate against.

Returns

A tuple of (error_string, validated_endpoint, realm, realm_key). The error_string will None if the rest of the values have been validated. The validated_endpoint will be the validated endpoint to sync to. The realm and realm_key will be set if validation was done through realms_conf.

swift.common.utils.whataremyips(ring_ip=None)

Get “our” IP addresses (“us” being the set of services configured by one *.conf file). If our REST listens on a specific address, return it. Otherwise, if listen on ‘0.0.0.0’ or ‘::’ return all addresses, including the loopback.

Parameters

ring_ip (str) – Optional ring_ip/bind_ip from a config file; may be IP address or hostname.

Returns

list of Strings of ip addresses

swift.common.utils.write_file(path, contents)

Write contents to file at path

Parameters
  • path – any path, subdirs will be created as needed

  • contents – data to write to file, will be converted to string

swift.common.utils.write_pickle(obj, dest, tmp=None, pickle_protocol=0)

Ensure that a pickle file gets written to disk. The file is first written to a tmp location, ensure it is synced to disk, then perform a move to its final location

Parameters
  • obj – python object to be pickled

  • dest – path of final destination file

  • tmp – path to tmp to use, defaults to None

  • pickle_protocol – protocol to pickle the obj with, defaults to 0

WSGI

WSGI tools for use with swift.

class swift.common.wsgi.ConfigDirLoader(conf_dir)

Bases: swift.common.wsgi.NamedConfigLoader

Read configuration from multiple files under the given path.

exception swift.common.wsgi.ConfigFileError

Bases: Exception

exception swift.common.wsgi.ConfigFilePortError

Bases: swift.common.wsgi.ConfigFileError

class swift.common.wsgi.ConfigString(config_string)

Bases: swift.common.wsgi.NamedConfigLoader

Wrap a raw config string up for paste.deploy.

If you give one of these to our loadcontext (e.g. give it to our appconfig) we’ll intercept it and get it routed to the right loader.

class swift.common.wsgi.NamedConfigLoader(filename)

Bases: paste.deploy.loadwsgi.ConfigLoader

Patch paste.deploy’s ConfigLoader so each context object will know what config section it came from.

class swift.common.wsgi.PipelineWrapper(context)

Bases: object

This class provides a number of utility methods for modifying the composition of a wsgi pipeline.

create_filter(entry_point_name)

Creates a context for a filter that can subsequently be added to a pipeline context.

Parameters

entry_point_name – entry point of the middleware (Swift only)

Returns

a filter context

index(entry_point_name)

Returns the first index of the given entry point name in the pipeline.

Raises ValueError if the given module is not in the pipeline.

insert_filter(ctx, index=0)

Inserts a filter module into the pipeline context.

Parameters
  • ctx – the context to be inserted

  • index – (optional) index at which filter should be inserted in the list of pipeline filters. Default is 0, which means the start of the pipeline.

startswith(entry_point_name)

Tests if the pipeline starts with the given entry point name.

Parameters

entry_point_name – entry point of middleware or app (Swift only)

Returns

True if entry_point_name is first in pipeline, False otherwise

class swift.common.wsgi.RestrictedGreenPool(size=1024)

Bases: eventlet.greenpool.GreenPool

Works the same as GreenPool, but if the size is specified as one, then the spawn_n() method will invoke waitall() before returning to prevent the caller from doing any other work (like calling accept()).

spawn_n(*args, **kwargs)

Create a greenthread to run the function, the same as spawn(). The difference is that spawn_n() returns None; the results of function are not retrievable.

class swift.common.wsgi.ServersPerPortStrategy(conf, logger, servers_per_port)

Bases: swift.common.wsgi.StrategyBase

WSGI server management strategy object for an object-server with one listen port per unique local port in the storage policy rings. The servers_per_port integer config setting determines how many workers are run per port.

Tracking data is a map like port -> [(pid, socket), ...].

Used in run_wsgi().

Parameters
  • conf (dict) – Server configuration dictionary.

  • logger – The server’s LogAdaptor object.

  • servers_per_port (int) – The number of workers to run per port.

iter_sockets()

Yields all known listen sockets.

log_sock_exit(sock, data)

Log a server’s exit.

loop_timeout()

Return timeout before checking for reloaded rings.

Returns

The time to wait for a child to exit before checking for reloaded rings (new ports).

new_worker_socks()

Yield a sequence of (socket, (port, server_idx)) tuples for each server which should be forked-off and started.

Any sockets for “orphaned” ports no longer in any ring will be closed (causing their associated workers to gracefully exit) after all new sockets have been yielded.

The server_idx item for each socket will passed into the log_sock_exit() and register_worker_start() methods.

no_fork_sock()

This strategy does not support running in the foreground.

register_worker_exit(pid)

Called when a worker has exited.

Parameters

pid (int) – The PID of the worker that exited.

register_worker_start(sock, data, pid)

Called when a new worker is started.

Parameters
  • sock (socket) – The listen socket for the worker just started.

  • data (tuple) – The socket’s (port, server_idx) as yielded by new_worker_socks().

  • pid (int) – The new worker process’ PID

class swift.common.wsgi.StrategyBase(conf, logger)

Bases: object

Some operations common to all strategy classes.

post_fork_hook()

Called in each forked-off child process, prior to starting the actual wsgi server, to perform any initialization such as drop privileges.

set_close_on_exec_on_listen_sockets()

Set the close-on-exec flag on any listen sockets.

shutdown_sockets()

Shutdown any listen sockets.

signal_ready()

Signal that the server is up and accepting connections.

class swift.common.wsgi.WSGIContext(wsgi_app)

Bases: object

This class provides a means to provide context (scope) for a middleware filter to have access to the wsgi start_response results like the request status and headers.

class swift.common.wsgi.WorkersStrategy(conf, logger)

Bases: swift.common.wsgi.StrategyBase

WSGI server management strategy object for a single bind port and listen socket shared by a configured number of forked-off workers.

Tracking data is a map of pid -> socket.

Used in run_wsgi().

Parameters
  • conf (dict) – Server configuration dictionary.

  • logger – The server’s LogAdaptor object.

iter_sockets()

Yields all known listen sockets.

log_sock_exit(sock, _unused)

Log a server’s exit.

Parameters
  • sock (socket) – The listen socket for the worker just started.

  • _unused – The socket’s opaque_data yielded by new_worker_socks().

loop_timeout()

We want to keep from busy-waiting, but we also need a non-None value so the main loop gets a chance to tell whether it should keep running or not (e.g. SIGHUP received).

So we return 0.5.

new_worker_socks()

Yield a sequence of (socket, opqaue_data) tuples for each server which should be forked-off and started.

The opaque_data item for each socket will passed into the log_sock_exit() and register_worker_start() methods where it will be ignored.

no_fork_sock()

Return a server listen socket if the server should run in the foreground (no fork).

register_worker_exit(pid)

Called when a worker has exited.

NOTE: a re-exec’ed server can reap the dead worker PIDs from the old server process that is being replaced as part of a service reload (SIGUSR1). So we need to be robust to getting some unknown PID here.

Parameters

pid (int) – The PID of the worker that exited.

register_worker_start(sock, _unused, pid)

Called when a new worker is started.

Parameters
  • sock (socket) – The listen socket for the worker just started.

  • _unused – The socket’s opaque_data yielded by new_worker_socks().

  • pid (int) – The new worker process’ PID

swift.common.wsgi.get_socket(conf)

Bind socket to bind ip:port in conf

Parameters

conf – Configuration dict to read settings from

Returns

a socket object as returned from socket.listen or ssl.wrap_socket if conf specifies cert_file

swift.common.wsgi.init_request_processor(conf_path, app_section, *args, **kwargs)

Loads common settings from conf Sets the logger Loads the request processor

Parameters
  • conf_path – Path to paste.deploy style configuration file/directory

  • app_section – App name from conf file to load config from

Returns

the loaded application entry point

Raises

ConfigFileError – Exception is raised for config file error

swift.common.wsgi.load_app_config(conf_file)

Read the app config section from a config file.

Parameters

conf_file – path to a config file

Returns

a dict

swift.common.wsgi.loadapp(conf_file, global_conf=None, allow_modify_pipeline=True)

Loads a context from a config file, and if the context is a pipeline then presents the app with the opportunity to modify the pipeline.

Parameters
  • conf_file – path to a config file

  • global_conf – a dict of options to update the loaded config. Options in global_conf will override those in conf_file except where the conf_file option is preceded by set.

  • allow_modify_pipeline – if True, and the context is a pipeline, and the loaded app has a modify_wsgi_pipeline property, then that property will be called before the pipeline is loaded.

Returns

the loaded app

swift.common.wsgi.make_env(env, method=None, path=None, agent='Swift', query_string=None, swift_source=None)

Returns a new fresh WSGI environment.

Parameters
  • env – The WSGI environment to base the new environment on.

  • method – The new REQUEST_METHOD or None to use the original.

  • path – The new path_info or none to use the original. path should NOT be quoted. When building a url, a Webob Request (in accordance with wsgi spec) will quote env[‘PATH_INFO’]. url += quote(environ[‘PATH_INFO’])

  • query_string – The new query_string or none to use the original. When building a url, a Webob Request will append the query string directly to the url. url += ‘?’ + env[‘QUERY_STRING’]

  • agent – The HTTP user agent to use; default ‘Swift’. You can put %(orig)s in the agent to have it replaced with the original env’s HTTP_USER_AGENT, such as ‘%(orig)s StaticWeb’. You also set agent to None to use the original env’s HTTP_USER_AGENT or ‘’ to have no HTTP_USER_AGENT.

  • swift_source – Used to mark the request as originating out of middleware. Will be logged in proxy logs.

Returns

Fresh WSGI environment.

swift.common.wsgi.make_pre_authed_env(env, method=None, path=None, agent='Swift', query_string=None, swift_source=None)

Same as make_env() but with preauthorization.

swift.common.wsgi.make_pre_authed_request(env, method=None, path=None, body=None, headers=None, agent='Swift', swift_source=None)

Same as make_subrequest() but with preauthorization.

swift.common.wsgi.make_subrequest(env, method=None, path=None, body=None, headers=None, agent='Swift', swift_source=None, make_env=<function make_env>)

Makes a new swob.Request based on the current env but with the parameters specified.

Parameters
  • env – The WSGI environment to base the new request on.

  • method – HTTP method of new request; default is from the original env.

  • path – HTTP path of new request; default is from the original env. path should be compatible with what you would send to Request.blank. path should be quoted and it can include a query string. for example: ‘/a%20space?unicode_str%E8%AA%9E=y%20es’

  • body – HTTP body of new request; empty by default.

  • headers – Extra HTTP headers of new request; None by default.

  • agent – The HTTP user agent to use; default ‘Swift’. You can put %(orig)s in the agent to have it replaced with the original env’s HTTP_USER_AGENT, such as ‘%(orig)s StaticWeb’. You also set agent to None to use the original env’s HTTP_USER_AGENT or ‘’ to have no HTTP_USER_AGENT.

  • swift_source – Used to mark the request as originating out of middleware. Will be logged in proxy logs.

  • make_env – make_subrequest calls this make_env to help build the swob.Request.

Returns

Fresh swob.Request object.

swift.common.wsgi.run_wsgi(conf_path, app_section, *args, **kwargs)

Runs the server according to some strategy. The default strategy runs a specified number of workers in pre-fork model. The object-server (only) may use a servers-per-port strategy if its config has a servers_per_port setting with a value greater than zero.

Parameters
  • conf_path – Path to paste.deploy style configuration file/directory

  • app_section – App name from conf file to load config from

  • allow_modify_pipeline – Boolean for whether the server should have an opportunity to change its own pipeline. Defaults to True

Returns

0 if successful, nonzero otherwise

swift.common.wsgi.wrap_conf_type(f)

Wrap a function whos first argument is a paste.deploy style config uri, such that you can pass it an un-adorned raw filesystem path (or config string) and the config directive (either config:, config_dir:, or config_str:) will be added automatically based on the type of entity (either a file or directory, or if no such entity on the file system - just a string) before passing it through to the paste.deploy function.

Storage Policy

class swift.common.storage_policy.BaseStoragePolicy(idx, name='', is_default=False, is_deprecated=False, object_ring=None, aliases='', diskfile_module='egg:swift#replication.fs')

Bases: object

Represents a storage policy. Not meant to be instantiated directly; implement a derived subclasses (e.g. StoragePolicy, ECStoragePolicy, etc) or use reload_storage_policies() to load POLICIES from swift.conf.

The object_ring property is lazy loaded once the service’s swift_dir is known via get_object_ring(), but it may be over-ridden via object_ring kwarg at create time for testing or actively loaded with load_ring().

add_name(name)

Adds an alias name to the storage policy. Shouldn’t be called directly from the storage policy but instead through the storage policy collection class, so lookups by name resolve correctly.

Parameters

name – a new alias for the storage policy

change_primary_name(name)

Changes the primary/default name of the policy to a specified name.

Parameters

name – a string name to replace the current primary name.

get_diskfile_manager(*args, **kwargs)

Return an instance of the diskfile manager class configured for this storage policy.

Parameters
  • args – positional args to pass to the diskfile manager constructor.

  • kwargs – keyword args to pass to the diskfile manager constructor.

Returns

A disk file manager instance.

get_info(config=False)

Return the info dict and conf file options for this policy.

Parameters

config – boolean, if True all config options are returned

load_ring(swift_dir, reload_time=None)

Load the ring for this policy immediately.

Parameters
  • swift_dir – path to rings

  • reload_time – time interval in seconds to check for a ring change

property quorum

Number of successful backend requests needed for the proxy to consider the client request successful.

classmethod register(policy_type)

Decorator for Storage Policy implementations to register their StoragePolicy class. This will also set the policy_type attribute on the registered implementation.

remove_name(name)

Removes an alias name from the storage policy. Shouldn’t be called directly from the storage policy but instead through the storage policy collection class, so lookups by name resolve correctly. If the name removed is the primary name then the next available alias will be adopted as the new primary name.

Parameters

name – a name assigned to the storage policy

validate_ring_data(ring_data)

Validation hook used when loading the ring; currently only used for EC

class swift.common.storage_policy.ECStoragePolicy(idx, name='', aliases='', is_default=False, is_deprecated=False, object_ring=None, diskfile_module='egg:swift#erasure_coding.fs', ec_segment_size=1048576, ec_type=None, ec_ndata=None, ec_nparity=None, ec_duplication_factor=1)

Bases: swift.common.storage_policy.BaseStoragePolicy

Represents a storage policy of type ‘erasure_coding’.

Not meant to be instantiated directly; use reload_storage_policies() to load POLICIES from swift.conf.

property ec_scheme_description

This short hand form of the important parts of the ec schema is stored in Object System Metadata on the EC Fragment Archives for debugging.

property fragment_size

Maximum length of a fragment, including header.

NB: a fragment archive is a sequence of 0 or more max-length fragments followed by one possibly-shorter fragment.

get_backend_index(node_index)

Backend index for PyECLib

Parameters

node_index – integer of node index

Returns

integer of actual fragment index. if param is not an integer, return None instead

get_info(config=False)

Return the info dict and conf file options for this policy.

Parameters

config – boolean, if True all config options are returned

property quorum

Number of successful backend requests needed for the proxy to consider the client PUT request successful.

The quorum size for EC policies defines the minimum number of data + parity elements required to be able to guarantee the desired fault tolerance, which is the number of data elements supplemented by the minimum number of parity elements required by the chosen erasure coding scheme.

For example, for Reed-Solomon, the minimum number parity elements required is 1, and thus the quorum_size requirement is ec_ndata + 1.

Given the number of parity elements required is not the same for every erasure coding scheme, consult PyECLib for min_parity_fragments_needed()

validate_ring_data(ring_data)

EC specific validation

Replica count check - we need _at_least_ (#data + #parity) replicas configured. Also if the replica count is larger than exactly that number there’s a non-zero risk of error for code that is considering the number of nodes in the primary list from the ring.

exception swift.common.storage_policy.PolicyError(msg, index=None)

Bases: ValueError

class swift.common.storage_policy.StoragePolicy(idx, name='', is_default=False, is_deprecated=False, object_ring=None, aliases='', diskfile_module='egg:swift#replication.fs')

Bases: swift.common.storage_policy.BaseStoragePolicy

Represents a storage policy of type ‘replication’. Default storage policy class unless otherwise overridden from swift.conf.

Not meant to be instantiated directly; use reload_storage_policies() to load POLICIES from swift.conf.

property quorum
Quorum concept in the replication case:

floor(number of replica / 2) + 1

class swift.common.storage_policy.StoragePolicyCollection(pols)

Bases: object

This class represents the collection of valid storage policies for the cluster and is instantiated as StoragePolicy objects are added to the collection when swift.conf is parsed by parse_storage_policies().

When a StoragePolicyCollection is created, the following validation is enforced:

  • If a policy with index 0 is not declared and no other policies defined, Swift will create one

  • The policy index must be a non-negative integer

  • If no policy is declared as the default and no other policies are defined, the policy with index 0 is set as the default

  • Policy indexes must be unique

  • Policy names are required

  • Policy names are case insensitive

  • Policy names must contain only letters, digits or a dash

  • Policy names must be unique

  • The policy name ‘Policy-0’ can only be used for the policy with index 0

  • If any policies are defined, exactly one policy must be declared default

  • Deprecated policies can not be declared the default

add_policy_alias(policy_index, *aliases)

Adds a new name or names to a policy

Parameters
  • policy_index – index of a policy in this policy collection.

  • aliases – arbitrary number of string policy names to add.

change_policy_primary_name(policy_index, new_name)

Changes the primary or default name of a policy. The new primary name can be an alias that already belongs to the policy or a completely new name.

Parameters
  • policy_index – index of a policy in this policy collection.

  • new_name – a string name to set as the new default name.

get_by_index(index)

Find a storage policy by its index.

An index of None will be treated as 0.

Parameters

index – numeric index of the storage policy

Returns

storage policy, or None if no such policy

get_by_name(name)

Find a storage policy by its name.

Parameters

name – name of the policy

Returns

storage policy, or None

get_object_ring(policy_idx, swift_dir)

Get the ring object to use to handle a request based on its policy.

An index of None will be treated as 0.

Parameters
  • policy_idx – policy index as defined in swift.conf

  • swift_dir – swift_dir used by the caller

Returns

appropriate ring object

get_policy_info()

Build info about policies for the /info endpoint

Returns

list of dicts containing relevant policy information

remove_policy_alias(*aliases)

Removes a name or names from a policy. If the name removed is the primary name then the next available alias will be adopted as the new primary name.

Parameters

aliases – arbitrary number of existing policy names to remove.

class swift.common.storage_policy.StoragePolicySingleton

Bases: object

An instance of this class is the primary interface to storage policies exposed as a module level global named POLICIES. This global reference wraps _POLICIES which is normally instantiated by parsing swift.conf and will result in an instance of StoragePolicyCollection.

You should never patch this instance directly, instead patch the module level _POLICIES instance so that swift code which imported POLICIES directly will reference the patched StoragePolicyCollection.

swift.common.storage_policy.get_policy_string(base, policy_or_index)

Helper function to construct a string from a base and the policy. Used to encode the policy index into either a file name or a directory name by various modules.

Parameters
  • base – the base string

  • policy_or_index – StoragePolicy instance, or an index (string or int), if None the legacy storage Policy-0 is assumed.

Returns

base name with policy index added

Raises

PolicyError – if no policy exists with the given policy_index

swift.common.storage_policy.parse_storage_policies(conf)

Parse storage policies in swift.conf - note that validation is done when the StoragePolicyCollection is instantiated.

Parameters

conf – ConfigParser parser object for swift.conf

swift.common.storage_policy.reload_storage_policies()

Reload POLICIES from swift.conf.

swift.common.storage_policy.split_policy_string(policy_string)

Helper function to convert a string representing a base and a policy. Used to decode the policy from either a file name or a directory name by various modules.

Parameters

policy_string – base name with policy index added

Raises

PolicyError – if given index does not map to a valid policy

Returns

a tuple, in the form (base, policy) where base is the base string and policy is the StoragePolicy instance for the index encoded in the policy_string.