delete_network kills dnsmasq on wrong machine
Bug #959794 reported by
JC Martin
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
When delete_network is invoked through nova-manage, the quantum driver is trying to delete the network resources on the machine where nova-manage is running, not where quantum is running.
This leads to gateway interfaces not cleaned up, and failure to kill dnsmasq.
Either the delete_network code should be invoked remotely, or nova-manage should not allow to delete network from outside of the the quantum node.
Changed in quantum: | |
status: | Fix Committed → Fix Released |
Changed in quantum: | |
milestone: | folsom-2 → 2012.2 |
To post a comment you must log in.
Thanks JC.
For Essex, I would say that you need to run nova-manage from the host running nova-network. To avoid the problem, I would think a simple check if self.host == network['host'], with an error printed if this is false, would work.
In Folsom, nova-network will likely go away all together when using Quantum, so there's no point in adding anything too fancy.