Cluster Membership

Concept

A Node can belong to at most one Cluster at any time. A node is referred to as an orphan node when it doesn’t belong to any cluster.

A node can be made a member of cluster when creation, or you can change the cluster membership after the cluster and the node have been created.

Listing Nodes in a Cluster

Using the command openstack cluster members list, you can list the nodes that are members of a specific cluster. For example, to list nodes in cluster c3, you can use the following command:

$ openstack cluster members list c3
+----------+--------+-------+--------+-------------+---------------------+
| id       | name   | index | status | physical_id | created_at          |
+----------+--------+-------+--------+-------------+---------------------+
| b28692a5 | stack1 | 1     | ACTIVE | fdf028a6    | 2015-07-07T05:23:40 |
| 4be10a88 | stack2 | 2     | ACTIVE | 7c87f545    | 2015-07-07T05:27:54 |
+----------+--------+-------+--------+-------------+---------------------+

You can use the name, the ID or the “short ID” of a cluster as the argument for node listing. If the specified cluster identifier cannot match any cluster or it matches more than one cluster, you will get an error message.

From the list, you can see the index, status, physical_id of each node in this cluster. Note that the id field and the physical_id field are shown as “short ID”s by default. If you want to see the full IDs, you can specify the --full-id option to indicate that:

$ openstack cluster members list --full-id c3
+------------...-+--------+-------+--------+-------------+-----------..-+
| id             | name   | index | status | physical_id | created_at   |
+------------...-+--------+-------+--------+-------------+-----------..-+
| b28692a5-25... | stack1 |  1    | ACTIVE | fdf0...     | 2015-07-07.. |
| 4be10a88-e3... | stack2 |  2    | ACTIVE | 7c87...     | 2015-07-07.. |
+------------...-+--------+-------+--------+-------------+-----------..-+

If the cluster size is very large, you may want to list the nodes in pages. This can be achieved by using the --marker option together with the --limit option. The marker option value specifies a node ID after which you want the resulted list to start; and the limit option value specifies the number of nodes you want to include in the resulted list. For example, the following command lists the nodes starting after a specific node ID with the length of the list set to 10:

$ openstack cluster members list --marker b28692a5 --limit 10 webservers

Another useful option for listing nodes is the --filters option. The option value accepts a string of format “K1=V1;K2=V2...”, where “K1” and “K2” are node properties for checking, “V1” and “V2” are values for filtering. The acceptable properties for filtering are name and status. For example, the following command lists cluster nodes from a cluster based on whether a node’s status is “ACTIVE”:

$ openstack cluster members list --filters status=ACTIVE webservers

Specify the Cluster When Creating a Node

There are several ways to make a node a member of a cluster. When creating a node using command openstack cluster node create, you can specify the option --cluster to tell Senlin to which cluster the new node belongs. Please refer to Nodes for detailed instructions.

Adding Node(s) to A Cluster

When you already have some nodes and some clusters, you can add some specified nodes to a specified cluster using the openstack cluster members add command. For example, the following command adds two nodes to a cluster:

$ openstack cluster members add --nodes node3,node4 cluster1

You can use the name, the ID or the “short ID” to name the node(s) to be added, you can also use the name, the ID or the “short ID” to specify the cluster. When the identifiers you specify cannot match any existing nodes or clusters respectively, you will receive an error message. If the identifier provided matches more than one object, you will get an error message as well.

Before Senlin engine performs the cluster membership changes, it will verify if the nodes to be added have the same profile type with the target cluster. If the profile types don’t match, you will get an error message.

When nodes are added to a cluster, they will get new index property values that can be used to uniquely identify them within the cluster.

Removing Node(s) from a Cluster

The openstack cluster command line also provides command cluster members del to remove node(s) from a cluster. In this case, you can use the name, the ID or the “short ID” to specify the node(s) and the cluster. The identifier specified must uniquely identifies a node or a cluster object, or else you will get an error message indicating that the request was rejected. The following command removes two nodes from a cluster:

$ openstack cluster members del --nodes node21,node22 webservers

When performing this operation, Senlin engine will check if the specified nodes are actually members of the specified cluster. If any node from the specified node list does not belong to the target cluster, you will get an error message and the command fails.

When nodes are removed from a cluster, they will get their index property reset to -1.

Replacing Node(s) in a Cluster

The openstack cluster command line also provides command cluster members replace to replace node(s) in a cluster. The argument “–nodes” is used to describe the list of node pairs like <OLD_NODE1=NEW_NODE1>. OLD_NODE is the name or ID of a node to be replaced, and NEW_NODE is the name or ID of a node as replacement. You can use the name, the ID or the “short ID” to specify the cluster. The identifier specified must uniquely identifies a node or a cluster object, or else you will get an error message indicating that the request was rejected. The following command replaces node21 with node22:

$ openstack cluster members replace --nodes node21=node22 webservers

When performing this operation, Senlin engine will check if the replaced nodes are actually members of the specified cluster. If any node from the specified node list does not belong to the target cluster, you will get an error message and the command fails.

When nodes are removed from the cluster, they will get their index property reset to -1.

See Also

Below are links to documents related to clusters and nodes: