openshift-on-openstack icon indicating copy to clipboard operation
openshift-on-openstack copied to clipboard

support for Aodh and Gnocchi alarms

Open ioggstream opened this issue 7 years ago • 4 comments

reproduce

  1. deploy on rhosp10 with autoscaling: true
  2. issue openstack alarm list

I expect

Alarm status to be in [ "ok", "alarm" ]

instead

Alarm status is "insufficient data"

my resolution

  • replace metadata on ServerGroup with the following.
          metadata: {"metering.server_group": {get_param: "OS::stack_id"}}
  • create Gnocchi + Aodh alarms like this
  cpu_alarm_high:
    type: OS::Aodh::GnocchiAggregationByResourcesAlarm
    properties:
      aggregation_method: mean
      metric: cpu_util
      comparison_operator: gt
      threshold: 50
      granularity: 600
      evaluation_periods: 1
      resource_type: instance
      enabled: {get_param: autoscaling}
      alarm_actions:
        - {get_attr: [scale_up_policy, alarm_url]}
      query:
        str_replace:
          template: '{"=": {"server_group": "stack_id"}}'
          params:
            stack_id: {get_param: "OS::stack_id"}

Feedback welcome!

Peace, R:

ioggstream avatar May 17 '17 08:05 ioggstream

Hi, For me, it's not good. No autoscalabity and I have always Alarm status is "insufficient data". Trace of aodh-evaluator.log. Always 0 statistics returned.

2017-07-04 19:06:18.921 16763 DEBUG aodh.evaluator [-] evaluating alarm d00d63bf-5a17-4eb0-b291-470138a9fb85 _evaluate_alarm /opt/stack/aodh/aodh/evaluator/init.py:257 2017-07-04 19:06:18.921 16763 DEBUG aodh.evaluator.threshold [-] query stats from 2017-07-04 19:04:18.921680 to 2017-07-04 19:06:18.921680 _bound_duration /opt/stack/aodh/aodh/evaluator/threshold.py:90 2017-07-04 19:06:18.949 16763 DEBUG aodh.evaluator.gnocchi [-] sanitize stats [] _sanitize /opt/stack/aodh/aodh/evaluator/gnocchi.py:48 2017-07-04 19:06:18.949 16763 DEBUG aodh.evaluator.gnocchi [-] pruned statistics to 0 _sanitize /opt/stack/aodh/aodh/evaluator/gnocchi.py:52

lborensky avatar Jul 04 '17 19:07 lborensky

There's an issue on RHOSP10 alarms.

Essentially they don't work out out of the box. You have to fix openstack before:

  • set default collection interval to 300 seconds
  • verify that every metric has 5 minute granularity

See

  • https://review.openstack.org/#/c/465499/
  • https://access.redhat.com/solutions/2953041

ioggstream avatar Jul 04 '17 22:07 ioggstream

@markllama ok, if it's the same as the proposed one you can close this ticket ;) @lborensky see related tickets for an official release

Related tickets:

  • Bug 1471234 - Unprivileged user can't access to its Gnocchi resources created by Ceilometer https://bugzilla.redhat.com/show_bug.cgi?id=1471234
  • https://review.openstack.org/#/c/482947/

ioggstream avatar Jul 28 '17 14:07 ioggstream

@lborensky fyi https://review.openstack.org/#/c/482947/ there will be an official release

ioggstream avatar Aug 28 '17 15:08 ioggstream