nova migrate does not update bridge configuration on destination node
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
New
|
Undecided
|
Unassigned |
Bug Description
This may be related to https:/
In Essex/stable, using 'nova migrate' to migrate an instance to another succeeds but the instance does not get a DHCP address on start because the dnsmasq entries for the instance do not get added to the dnsmasq config file.
To duplicate:
1) boot instance
2) on the compute node where the instance is running view the dnsmasq config file (in this case /var/lib/
3) run 'nova migrate' to migrate the instance.
4) on the original source compute node verify that the mac address is still present in the config file
5) on the compute node where the instance is now running verify that the mac address does not exist in the dnsmasq config file
As a result the instance does not get a response from the DHCPDISCOVER.
This can be resolved by restarting nova-network on the source and destination compute nodes and rebooting the instance.
This is present in Essex/stable - I have not verified it in newer releases or trunk.
Changed in nova: | |
status: | Incomplete → New |
(In case it's not obvious to everyone, this is with multi_host mode)
We expect this to be regenerating the dnsmasq config:
+ # NOTE(vish): this is necessary to update dhcp api.setup_ networks_ on_host( context,
+ self.network_
+ instance_ref,
+ self.host)
this is called after the instance host field is updated, so you'd expect the dnsmasq config to be regenerated correctly - i.e. linux_net. update_ dhcp() should be called and get_dhcp_hosts() should return the instance on the destination host but not on the source note
that code is from stable/essex and was part of the fix for bug #939060
Hard to come to any conclusions without logs. Could you attach compute and network logs from the source and destination hosts?