node id is not recorded if provisioning does not complete
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cue |
Fix Committed
|
Medium
|
Unassigned |
Bug Description
Node ID is not being recorded if a broker does not fully provision. This is an issue as trying to delete a cluster while it's building, or if the worker is interrupted, or if there is some failure such that the cluster is not rolled back during a cluster create, the node cannot be deleted along with the cluster.
Also, if the node ID does not exist Cue is issuing a nova delete with None as the ID, which causes an exception in novaclient.
Reproduction steps:
- create a cluster of size 3
- kill cue-worker
- launch zkCli
- rmr /cue/taskflow/jobs
- exit zkCli
- start cue-worker
- cluster status should stay in BUILDING after worker has consumed the job
- issue a cluster delete
- nova client will throw exceptions stating instance of type None cannot be found
Changed in cue: | |
importance: | Undecided → Medium |
description: | updated |
description: | updated |
Reviewed: https:/ /review. openstack. org/190014 /git.openstack. org/cgit/ stackforge/ cue/commit/ ?id=a8a67b20071 9fc6bd5315fe739 9b9119198fa1a9
Committed: https:/
Submitter: Jenkins
Branch: master
commit a8a67b200719fc6 bd5315fe7399b91 19198fa1a9
Author: Min Pae <email address hidden>
Date: Tue Jun 9 21:07:20 2015 -0700
Moving Node update earlier in cluster create flow
Moving Node update to happen earlier in the cluster create process, such
that node VM ID is recorded regardless of cluster create success or
failure so that any VMs created during a failed cluster create can be
properly destroyed.
Change-Id: I0c46e18986c615 948300b145a9e68 bc6607e13ec
Closes-Bug: 1463605