ceilometer-compute service terminates in case of exception

[Stefano Zilli]

In case of exception during the poll_and_publish call ceilometer-compute service terminates. It should be better to manage the exception and keep the service alive as an outage of another service, like keystone, could lead to the death of all the running ceilometer agents.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/46203

[Harri Hämäläinen (hhamalai)]

Pollsters are handling exceptions as should, but nova_client can still raise an exception which will finally cause the termination of Ceilometer compute agent. This might happen as said due to an outage of another service or temporary fault in system.

The abstract problem here is how Ceilometer should react in this case. No measurements can be done on host if Ceilometer cannot query the host information. On the other hand no measurements can be done if the Ceilometer compute agent is down either, so I prefer the persistency of the Ceilometer compute agent and logging of the error condition.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/46203
Committed: http://github.com/openstack/ceilometer/commit/0e01da763ab7b3783ae86aa6582c38ebdc6e2378
Submitter: Jenkins
Branch: master

commit 0e01da763ab7b3783ae86aa6582c38ebdc6e2378
Author: Harri Hämäläinen <email address hidden>
Date: Thu Sep 12 09:34:34 2013 +0300

    Catch exceptions from nova client in poll_and_publish

    Ceilometer compute agent dies if nova client raises an exception while it is
    retrieving server instances. This might happen e.g. when some OpenStack API is
    temporarily unavailable

    Fixes LP Bug #1218889

    Change-Id: I808dcfae18d23240f8e095d6c97c8dede7dede8f

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/grizzly
Review: https://review.openstack.org/46793

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/46793
Committed: http://github.com/openstack/ceilometer/commit/b2fe778c66d06d25f515cbbcb7da61939d68fc28
Submitter: Jenkins
Branch: stable/grizzly

commit b2fe778c66d06d25f515cbbcb7da61939d68fc28
Author: Harri Hämäläinen <email address hidden>
Date: Thu Sep 12 09:34:34 2013 +0300

    Catch exceptions from nova client in poll_and_publish

    Ceilometer compute agent dies if nova client raises an exception while it is
    retrieving server instances. This might happen e.g. when some OpenStack API is
    temporarily unavailable

    Conflicts:
     tests/agentbase.py
     tests/compute/test_manager.py

    Fixes LP Bug #1218889
    Change-Id: I808dcfae18d23240f8e095d6c97c8dede7dede8f
    (cherry picked from commit 0e01da763ab7b3783ae86aa6582c38ebdc6e2378)