| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Add SCTP protocol scenario tests https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/738643 | 07:55 |
|---|---|---|
| playersons | Hi guys, can I bother anyone with an issue I am having with Octavia? I am new here and not familiar with the rules, so apologies upfront. | 11:03 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Add SCTP protocol scenario tests https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/738643 | 11:18 |
| playersons | I am struggling with one of three lb ports/control nodes that won't turn ACTIVE as the other two. This is an OSISM kolla-ansible setup that I have very little experience with. I'd appreciate any hints where to look for misconfigurations. Thanks! | 12:22 |
| playersons | ? | 12:34 |
| gthiemonge | playersons: hi | 13:06 |
| gthiemonge | do you have running LBs on this deployment? there are o-hm0 interfaces on each controller, those interfaces are plugged into the lb-mgmt-net tenant network that was created for octavia, you should see incoming traffic with tcpdump on o-hm on the 3 controllers (LBs should send UDP heartbeat packets) | 13:08 |
| playersons | I can see the o-hm0 interfaces in each controller, and they are listed in the lb-mgmt-net, but one is staying down, even if I manually remove and add it back with ovs-vsctl | 13:46 |
| playersons | found a difference in the openvswitch_vswitchd containers, this one is an ACTIVE port: | 14:16 |
| playersons | Port vxlan-ac106644 Interface vxlan-ac106644 type: vxlan options: {df_default="true", egress_pkt_mark="0", in_key=flow, local_ip="172.16.102.67", out_key=flow, remote_ip="172.16.102.68"} | 14:17 |
| playersons | while the Bridge br-tun part on the DOWN node does not contain anything vxlan related | 14:18 |
| johnsom | playersons: That sounds kolla setup related. I don’t know much about kolla and how it configured the controllers. Have you tried asking in the #openstack-kolla channel? | 14:59 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Add SCTP protocol scenario tests https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/738643 | 14:59 |
| johnsom | I would assume if one controller is setup for vxlan, they all should be | 15:00 |
| *** dkehn_ is now known as dkehn | 15:02 | |
| servagem_ | Hi | 17:24 |
| servagem_ | I have a openstack with 3 AZ with the same network (L2). I would like to have high availability for loadbalancers. I mean, if LB's amphora are running in on AZ and the AZ fails, the LB failover for others AZ. | 17:24 |
| servagem_ | Octavia is configured with active_standby. Is it possible to configure octavia with AZ anti-affinity (force amphora master and backup to be deployed in different AZs) ? | 17:25 |
| johnsom | servagem_ Not currently. Nova doesn't provide cross AZ anti-affinity filters (server groups) and we have not implemented this yet. There are some proposed patches, but they are not finished or merged. | 17:27 |
| servagem_ | ok, and if I create a LB without specifying the AZ, it can be deployed in any AZ. In case of AZ failure (the one that has the amphora), would LB failover to other AZs? | 17:30 |
| johnsom | No, LBs will always be created and re-created in the same AZ. If you don't specify an AZ it will use the nova default | 17:31 |
| servagem_ | ok, I think one alternative would be create a host aggregate with all compute of those AZ and configure the octavia flavor to filter for that aggregate. Would it work? | 17:34 |
| johnsom | You could try, but I don't think nova will allow it. | 17:35 |
| servagem_ | johnsom: ok. Thank you! | 17:36 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!