| skraynev | Hi folks, I am just wondering: is this story valid or was rejected: https://storyboard.openstack.org/#!/story/2009691 ? | 12:16 |
|---|---|---|
| tweining | at first glance I would say it is a valid bug | 12:18 |
| skraynev | hm. in this case it looks like a "missed" story, I mean, that it was created a long time ago, but has not any comments. I did not find any updates about it, I mean | 12:21 |
| tweining | that said I don't know if using Octavia with two subnets is a supported feature or not | 12:23 |
| tweining | hm, looking more closely it looks like the loadbalancer was configured to use lb-subnet-2 only, in which case this behavior would make sense. So I guess it is really up to the question whether two subnets are supported or not | 12:25 |
| skraynev | hm. I though about the scenario, when LB use small subnet and during failover it has not enough IP addresses for creation one more amphora with IPs. So my thought was about some logic, which allow to do failover in different order: deallocate IP from old amphora and allocate to new one. | 12:29 |
| skraynev | I understand, that it's less safe, but I don't see another way to solve this situation except re-create Lb with correct subnet. | 12:30 |
| tweining | yeah, I think you're right | 12:31 |
| skraynev | I missed, that in task mentioned about another subnets, sorry for that. My corner case - one subnet and it has not more IPs | 12:31 |
| gthiemonge | Hi, I don't know why we didn't reply in this story :/ | 12:32 |
| gthiemonge | looks valid | 12:32 |
| gthiemonge | I don't think that deallocating the old IP is a good idea, it would interrupt the data plane traffic | 12:34 |
| gthiemonge | I think we need to valid that a failover is possible (enough IP addresses are available) in the API, we do a similar validation when a LB is created: https://opendev.org/openstack/octavia/src/branch/master/octavia/api/v2/controllers/load_balancer.py#L149-L164 | 12:35 |
| gthiemonge | this way, we would have a clear error message | 12:36 |
| kevko | hi, is it possible with octavia amphora ...start amphoras always on separate hosts ? lets say on separate compute hosts in two differents AZs ? | 13:37 |
| gthiemonge | kevko: hi, ATM it's possible to start the amphorae in an ACTIVE-STANDBY LB in separate compute nodes but in the same AZ | 14:04 |
| gthiemonge | kevko: there's a proposal for creating the amphorae in different AZs: https://review.opendev.org/c/openstack/octavia/+/558962 | 14:04 |
| gthiemonge | kevko: we hope that it will be included in the Bobcat release | 14:04 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: DNM/WIP Testing server output https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/877667 | 15:07 |
| johnsom | kevko Yes, https://github.com/openstack/octavia/blob/master/octavia/common/config.py#L702 | 16:07 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: DNM/WIP Testing server output https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/877667 | 18:05 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!