Cannot colocate services on machines running neutron-contrail
Bug #1395860 reported by
Greg Lutostanski
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron-api (Juju Charms Collection) |
Fix Released
|
Undecided
|
Robert Ayres |
Bug Description
When running neutron-contrail in "density" (smoosh) mode where services are co-located onto a nova-compute node when adding the virtual router onto the compute node other services (inside lxcs) are disconnected from the network.
Networking up until that point is fine -- the call to "contrail-
After this occurs in a density deployment the lxcs are still running but there are no internal bridges setup -- just the interfaces of the physical machines; so they are effectively cut off from the network.
Related branches
affects: | charms → neutron-api (Juju Charms Collection) |
To post a comment you must log in.
When the neutron-contrail charm sets up the vrouter vhost0 interface, it will ifdown (and subsequently ifup) the default route interface. In this case the interface will be the lxc bridge (br0/lxcbr0). I think probably what's happening is the lxc veth interfaces aren't being reattached to the bridge when it's brought back up. 'brctl addif <bridge> <veth>' would probably make the said containers work again.
The charm can probably be worked so it takes a note of bridge interfaces to reattach when it brings the bridge back up. A simple workaround now is to deploy nova-compute/ neutron- contrail services first and then lxc ones.