Jump to: navigation, search


Note: the current cloud watch implementation is a minimal implementation, and does not provide feature parity with AWS Cloudwatch. It is primarily required to enable metric collection for HA and Autoscaling functionality, but can also be used to configure user defined alarms, as outlined below. This feature will be deprecated or removed during the Havana cycle as we move to using Ceilometer as a metric/alarm service instead.

To use cloud watch you need the following in your template:

    "MyOwnAlarm": {
     "Type": "AWS::CloudWatch::Alarm",
     "Properties": {
        "AlarmDescription": "Restart the WikiDatabase if httpd fails > 3 times in 10 minutes",
        "MetricName": "ServiceFailure",
        "Namespace": "system/linux",
        "Statistic": "SampleCount",
        "Period": "300",
        "EvaluationPeriods": "1",
        "Threshold": "2",
        "AlarmActions": [ { "Ref": "WebServerRestartPolicy" } ],
        "ComparisonOperator": "GreaterThanThreshold"

Note: Currently the only Action that is supported is HEAT::HA::Restarter

Then you need to call this from in your vm: /opt/aws/bin/cfn-push-stats --watch MyOwnAlarm [other args]

To see this all together look at the HA template

CloudWatch API

An AWS-compatible API to the heat CloudWatch alarms and metrics is in-progress. Not all AWS actions are supported (yet), but you can do the following:

  • List Alarms
  • List Metrics (related to alarms only, generic Metric datapoints not yet supported)
  • Create Metric Datapoints (similar to current cfn-push-stats functionality but authenticated)
  • Set Alarm state (temporarily, useful for testing alarm actions)

Status of API actions as they are implemented will be tracked in the API roadmap page(https://github.com/heat-api/heat/wiki/Roadmap-Feature:-cloudwatch-api)

You can use the Cloudwatch API by doing the following:

heat-api-cloudwatch -d &
tail -f /var/log/heat/api-cloudwatch.log

Note that it is not currently necessary to run heat-api-cloudwatch for HA and autoscaling to work correctly - instances still report Cloudwatch metric data via the metadata server (for the time being at least..)

CloudWatch Client "**heat-watch**"

A client has been created to demonstrate the CloudWatch API - it is **heat-watch**

Usage examples:

List Alarms:

heat-watch describe

List Metrics Datapoints:

heat-watch metric-list

Create Metric Datapoints:

heat-watch metric-put-data HttpFailureAlarm system/linux ServiceFailure Count 1

Set Alarm state:

heat-watch -d set-state HttpFailureAlarm ALARM