Apmec Policy Framework¶

Introduction¶

Apmec policy actions framework provides the MEC operators and MEA vendors to write a pluggable action that manages their own MEAs. Currently Apmec already provided some common actions like autoscaling, respawning, and logging. With this framework the custom actions can be easily applied for the management purpose.

How to write a new policy action¶

A policy action for apmec is a python module which contains a class that inherits from “apmec.mem.policy_actions.abstract_action.AbstractPolicyAction”. If the driver depends/imports more than one module, then create a new python package under apmec/mem/policy_actions folder. After this we have to mention our driver path in setup.cfg file in root directory.

For example:

apmec.apmec.policy.actions =
  respawn = apmec.mem.policy_actions.respawn.respawn:MEAActionRespawn

Following methods need to be overridden in the new action:

def get_type(self)

This method must return the type of action. ex: respawn

def get_name(self)

This method must return the symbolic name of the mea policy action.

def get_description(self)

This method must return the description for the policy action.

def execute_action(self, plugin, context, mea, arguments)

This method must expose what will be executed with the policy action. ‘arguments’ is used to add more options for policy actions. For example, if action is scaling, ‘arguments’ should let you know ‘scaling-out’ or ‘scaling-in’ will be applied.

Event and Auditing support¶

This function can be used to describe the execution process of policy. For example:

_log_monitor_events(context, mea_dict, "ActionRespawnHeat invoked")

How to combine policy with existing monitoring framework in Apmec¶

In the monitoring policy section, you can specify the monitors details with corresponding action.

The below example shows how policy is used for alarm monitor. Example Template —————-

policies:
    - vdu1_cpu_usage_monitoring_policy:
        type: tosca.policies.apmec.Alarming
        triggers:
          resize_compute:
              event_type:
                  type: tosca.events.resource.utilization
                  implementation: ceilometer
              metrics: cpu_util
              condition:
                  threshold: 50
                  constraint: utilization greater_than 50%
                  period: 65
                  evaluations: 1
                  method: avg
                  comparison_operator: gt
              actions: [respawn]