nova-network gets release_fixed_ip events from someplace, but the database still keeps them associated with instances
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Triaged
|
High
|
Unassigned | ||
nova (Ubuntu) |
Triaged
|
Medium
|
Unassigned |
Bug Description
We've been seeing a lot of instances simply vanish from the network. Usually people have been willing to work around this by simply rebooting or re-creating their instances, but it's troubling for long-running instances (especially those that have volumes associated).
Here's the relevant bit of nova-network.log for one of these:
2012-07-16 14:06:32 DEBUG nova.rpc.amqp [-] received {u'_context_roles': [u'admin'], u'_context_
2012-07-16 14:06:32 DEBUG nova.rpc.amqp [req-d0905711-
2012-07-16 14:06:32 DEBUG nova.network.
Then the dhcpbridge shows it being revoked:
2012-07-16 14:04:29 DEBUG nova.dhcpbridge [-] Called 'old' for mac 'fa:16:3e:11:c5:37' with ip '10.55.60.141' from (pid=23699) main /usr/bin/
2012-07-16 14:06:32 DEBUG nova.dhcpbridge [-] Called 'del' for mac 'fa:16:3e:11:c5:37' with ip '10.55.60.141' from (pid=24946) main /usr/bin/
Is there any way we can find out what might have placed the release_fixed_ip event on the message queue? There doesn't seeem to be any other mention of the IP in the nova logs on any of our systems.
Changed in nova: | |
status: | Incomplete → Confirmed |
tags: | added: libvirt |
description: | updated |
Changed in nova (Ubuntu): | |
status: | Confirmed → Triaged |
importance: | Undecided → Medium |
I should probably clarify that there are no API requests matching this release_fixed_ip event, and the instance owner explicitly did not want the event to take place.