failover takes 15-20 minutes and the dashboard is slow and it take longer time to launch vm

[Moshe Levi]

we have HA environment with centos

3 controller
3 Compute
1 storage

- neutron with VLANs
- Mellanox plugin + SR-IOV
- Cinder + iSER

After failover we were able to see the VMs in the dashboard after 15-20 min.
also before the failover the VM's assigned with floating ip and were accessible via external network.

after failover we couldn't access the vm from external network.

fuel iso fuel-community-master-52-2014-08-04_05-27-27.iso

attached snapshot

[Moshe Levi (moshele)]

also vm didn't get private ip

[Tomasz 'Zen' Napierala (tzn)]

Please let us know how do you trigger failover

[Moshe Levi (moshele)]

we poweroff the active controller

[Moshe Levi (moshele)]

what iso fix this issue?

[Tomasz 'Zen' Napierala (tzn)]

This does not look liek duplicate. First - completely different usage scanario, like neutron instead of nova-network. Second, no traces in the logs. Third - I don't suspect regression here, we are on 5.1

[Moshe Levi (moshele)]

I test again on new setup the vm get the ip address so that not an issue.

the only problem is that the dashboard is slow and it take longer time to launch vm

[Tomasz 'Zen' Napierala (tzn)]

Thanks Moshe!
Would you please attach new diag snapshot from the current deployment? It will be clearer to read.

[Moshe Levi (moshele)]

the snapshot is attached

[Moshe Levi (moshele)]

the dashboard I mean the openstack dashboard

[Moshe Levi (moshele)]

we reproduce the issue without mellanox hardware
we used HP switch

also artem_panchenko in fuel qa say he seem the same issue
and it related to this bug https://bugs.launchpad.net/fuel/+bug/1340657

<artem_panchenko> moshele_ this is because of keystone becomes slowly after failover, here is the bug: https://bugs.launchpad.net/fuel/+bug/1340657

[Dmitry Borodaenko (angdraug)]

Yes, this is duplicate of bug #1340657.