RDO tickets are now tracked in Jira https://issues.redhat.com/projects/RDO/issues/
Bug 1049369 - new ceilometer-agent-notification service needs to be started alongside the other ceilometer services
Summary: new ceilometer-agent-notification service needs to be started alongside the o...
Keywords:
Status: CLOSED WORKSFORME
Alias: None
Product: RDO
Classification: Community
Component: openstack-puppet-modules
Version: unspecified
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
: Icehouse
Assignee: Francesco Vollero
QA Contact: Ami Jeain
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2014-01-07 13:39 UTC by Eoghan Glynn
Modified: 2014-11-21 16:52 UTC (History)
6 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2014-11-21 16:52:56 UTC
Embargoed:


Attachments (Terms of Use)

Description Eoghan Glynn 2014-01-07 13:39:19 UTC
Description of problem:

For icehouse-1, the ceilometer collector service was split so as to separate the logic for consuming metering messages emitted by the ceilometer compute/central agents and the notifications emitted by the base openstack services:

  https://blueprints.launchpad.net/ceilometer/+spec/split-collector

As a result, a new opestack-ceilometer-agent-notification service has been added to the openstack-ceilometer-collector package.

However this service is not being launched during a packstack run, hence any meters derived from openstack notifications are not being persisted.


Version-Release number of selected component (if applicable):

openstack-packstack-2013.2.1-0.27.dev936.fc21.noarch


How reproducible:

100%


Steps to Reproduce:

0. Install packstack --allinone in the usual way

1. Check that the openstack-ceilometer-agent-notification is not running:

  $ sudo service openstack-ceilometer-agent-notification status

2. Spin up a tiny instance via nova boot

3. Check immediately whether the instance meter is for the new resource is recorded:

  $ sleep 5 ; ceilometer meter-list -m instance -q resource=$INSTANCE_ID

4. Wait for the ceilometer compute agent polling cycle to roll around, at which point the instance meter for the new VM should be available:

  $ INTERVAL=$(sudo grep -A1 meter_pipeline /etc/ceilometer/pipeline.yaml | grep interval | cut -f2 -d: | tr -d ' ')
  $ sleep $INTERVAL ; ceilometer meter-list -m instance -q resource=$INSTANCE_ID


Actual results:

The openstack-ceilometer-agent-notification service is not running at step #1 and the instance meter is not seen at step #3 above.


Expected results:

The openstack-ceilometer-agent-notification service should be running at step #1 and the instance meter should be seen at step #3 above.


Additional info:

A workaround is to explicitly start the new service immediately after the packstack run completes:

  $ sudo service openstack-ceilometer-agent-notification start

Comment 1 Kevin Whitney 2014-01-07 14:08:37 UTC
Note for tester clarification.  

1) The new service is not known to "chkconfig" and will need to be restarted manually upon reboot of host.

2) A side-effect is that openstack-status does not show the agent-notifier service at all


Note You need to log in before you can comment on or make changes to this bug.