Back to bug 1324739

Who When What Removed Added
Red Hat Bugzilla Rules Engine 2016-04-07 07:27:32 UTC Target Release --- 8.0
Marius Cornea 2016-04-07 10:39:24 UTC Status NEW CLOSED
Resolution --- DUPLICATE
Last Closed 2016-04-07 06:39:24 UTC
Marius Cornea 2016-04-08 14:44:27 UTC Status CLOSED NEW
Resolution DUPLICATE ---
Keywords Reopened
Marios Andreou 2016-04-11 09:48:42 UTC CC mandreou
Mike Burns 2016-04-11 13:36:33 UTC CC sasha
Mike Burns 2016-04-11 14:03:50 UTC Priority unspecified urgent
Target Milestone --- ga
Mike Burns 2016-04-11 14:04:14 UTC Assignee athomas mandreou
Giulio Fidente 2016-04-11 21:06:19 UTC CC gfidente
Mike Burns 2016-04-12 00:04:30 UTC Status NEW ON_DEV
Component rhel-osp-director openstack-tripleo-heat-templates
Link ID OpenStack gerrit 304342
Doc Text Cause: A change in hostname values for compute nodes

Consequence: Some issues with nova hostnames will result possibly causing anything from nova commands not working to issues controlling old guests. Extra incorrect information will be seen in nova hypervisor lists as well

Workaround (if any): When running any upgrades or subsequent scaling operations, include an environment file like this:

parameter_defaults
ComputeHostnameFormat: '%stackname%-compute-%index%'

Result: The existing naming convention will be maintained and persist after upgrade.
Target Milestone ga async
Doc Type Bug Fix Known Issue
Steve Linabery 2016-04-15 20:37:59 UTC Status ON_DEV MODIFIED
CC slinaber
Fixed In Version openstack-tripleo-heat-templates-0.8.14-9.el7ost
errata-xmlrpc 2016-04-15 20:55:34 UTC Status MODIFIED ON_QA
Arik Chernetsky 2016-04-17 09:17:28 UTC QA Contact achernet mcornea
Marius Cornea 2016-04-18 16:43:49 UTC Status ON_QA VERIFIED
Andrew Dahms 2016-04-19 00:02:35 UTC CC adahms
Doc Text Cause: A change in hostname values for compute nodes

Consequence: Some issues with nova hostnames will result possibly causing anything from nova commands not working to issues controlling old guests. Extra incorrect information will be seen in nova hypervisor lists as well

Workaround (if any): When running any upgrades or subsequent scaling operations, include an environment file like this:

parameter_defaults
ComputeHostnameFormat: '%stackname%-compute-%index%'

Result: The existing naming convention will be maintained and persist after upgrade.
Currently, issues with nova host names can potentially cause a number of issues such as nova commands not working, or issues controlling old guests. Additional incorrect information is also seen in nova hypervisor lists. As a workaround, when running any upgrades or subsequent scaling operations, include an environment file similar to the following:

parameter_defaults
ComputeHostnameFormat: '%stackname%-compute-%index%'

The existing naming convention is maintained, and persists after upgrade.
Scott Lewis 2016-04-19 12:42:15 UTC CC sclewis
Flags needinfo?(mandreou)
Marios Andreou 2016-04-19 15:46:34 UTC Doc Type Known Issue Bug Fix
Flags needinfo?(mandreou)
Red Hat Bugzilla 2016-04-19 15:46:34 UTC Doc Type Bug Fix Known Issue
Scott Lewis 2016-04-19 15:50:09 UTC Flags needinfo?(adahms)
Mike Burns 2016-04-19 15:51:40 UTC Doc Text Currently, issues with nova host names can potentially cause a number of issues such as nova commands not working, or issues controlling old guests. Additional incorrect information is also seen in nova hypervisor lists. As a workaround, when running any upgrades or subsequent scaling operations, include an environment file similar to the following:

parameter_defaults
ComputeHostnameFormat: '%stackname%-compute-%index%'

The existing naming convention is maintained, and persists after upgrade.
Cause: There is divergence in the default value for the ComputeHostnameFormat parameter (https://github.com/openstack/tripleo-heat-templates/blob/a12087715f0fe4251a95ab67120023d553c24a45/overcloud.yaml#L790 ) between what was originally deployed for RHOS 7.3 and what we now have in RHOS 8.0 ((overcloud-compute-N" vs "overcloud-novacompute-N")

Consequence: This causes the upgraded compute nodes to have an incorrect setting for hostname, ultimately resulting in a duplicate listing for the node in the nova hypervisor-list amongst other problems, including issues restarting existant/pre-upgrade vms.

Fix: Preserve the ComputeHostnameFormat to its existing value during the ugprade so it doesn't change from the overcloud-compute-N pattern.

Result: The compute nodes are upgraded correctly and in particular keep their existing, pre-upgrade hostname.
Doc Type Known Issue Bug Fix
Andrew Dahms 2016-04-19 16:00:59 UTC Doc Text Cause: There is divergence in the default value for the ComputeHostnameFormat parameter (https://github.com/openstack/tripleo-heat-templates/blob/a12087715f0fe4251a95ab67120023d553c24a45/overcloud.yaml#L790 ) between what was originally deployed for RHOS 7.3 and what we now have in RHOS 8.0 ((overcloud-compute-N" vs "overcloud-novacompute-N")

Consequence: This causes the upgraded compute nodes to have an incorrect setting for hostname, ultimately resulting in a duplicate listing for the node in the nova hypervisor-list amongst other problems, including issues restarting existant/pre-upgrade vms.

Fix: Preserve the ComputeHostnameFormat to its existing value during the ugprade so it doesn't change from the overcloud-compute-N pattern.

Result: The compute nodes are upgraded correctly and in particular keep their existing, pre-upgrade hostname.
Previously, there was a divergence in the default value for the ComputeHostnameFormat parameter between what was originally deployed for Red Hat Enterprise Linux OpenStack Platform 7.3 and the current default value in Red Hat OpenStack Platform 8.0 ("overcloud-compute-N" instead of "overcloud-novacompute-N"). This would cause upgraded compute nodes to have an incorrect setting for their host name, resulting in issues such as the node being listed twice in the output of 'nova hypervisor-list', and issues restarting existing instances. With this update, the value of ComputeHostnameFormat is now preserved during upgrade so that it does not change from the overcloud-compute-N pattern, allowing compute nodes to be upgraded correctly and to retain the same host name as that prior to the upgrade.
Flags needinfo?(adahms)
errata-xmlrpc 2016-04-20 00:29:47 UTC Status VERIFIED RELEASE_PENDING
errata-xmlrpc 2016-04-20 13:04:44 UTC Status RELEASE_PENDING CLOSED
Resolution --- ERRATA
Last Closed 2016-04-07 06:39:24 UTC 2016-04-20 09:04:44 UTC

Back to bug 1324739