Can't delete instance if nova-network isn't responding

Bug #985185 reported by Joe Harnish
24
This bug affects 4 people
Affects Status Importance Assigned to Milestone
OpenStack Compute (nova)
Invalid
Medium
Unassigned

Bug Description

This is a fresh install of Ubuntu 12.04 beta2, with 2012.1 Nova. Following steps defined here: http://www.hastexo.com/resources/docs/installing-openstack-essex-4-ubuntu-1204-precise-pangolin

when I issued the nova boot command the build failed during the network phase. Now the instance is setting in an error state. I run the nova delete command it sits there.

Nothing new appears in the nova-api.log file.

nova-network is up and running.

Revision history for this message
Ying Chun Guo (daisy-ycguo) wrote :

Sometimes, I will meet the same error. I guess, it may be caused by the asynchronous invocation. "nova delete" will return success before the instance is actually deleted.

Revision history for this message
Vish Ishaya (vishvananda) wrote :

I can only replicate this if the network worker is not recponding. If the network worker is down, you cannot terminate an instance properly. This is because compute attempts to get the network information from the network worker so it can tear down the network if it fails to communicate with the network worker.

I'm not sure if we have a good solution for this except for hiding terminated instances from the user immediately. I don't think it is safe to actually delete the vm until we can communicate with the network worker. This may also be partially resolved by the move of the relevant code up into the api.

Changed in nova:
status: New → Triaged
importance: Undecided → Medium
summary: - Can't delete Instance after Error state
+ Can't delete instance if nova-network isn't responding
Revision history for this message
Mark Russell (marrusl) wrote :

The trouble is if someone deletes the network with nova-manage while an instance is still running in it.

Then it seems the only way to terminate is create a new network, edit the DB to match the original network ID, and then terminate.

Revision history for this message
Vish Ishaya (vishvananda) wrote :

I guess the simple answer is don't do that! I suppose we could do a consistency check in nova-manage network delete to be sure there are no vms in that network before allowing them to be deleted. Really though deletes are never permanent, you could just set deleted=0 for the old network and its fixed ips in the db instead of creating a new one.

Revision history for this message
Justin Shepherd (jshepher) wrote :

Marking this issue as invalid, as it is filed against Essex, and has not been updated in almost two years. If you believe this to be in error, please update this issue.

Changed in nova:
status: Triaged → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.