| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Fix request_interval uses https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/759357 | 07:43 |
|---|---|---|
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Add a test for custom compute flavors for amphorae https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/773696 | 08:03 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia-tempest-plugin master: Fix request_interval uses https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/759357 | 11:36 |
| mephmanx | johnsom Good morning! Happy monday! I did quite a bit of cleanup over the weekend and I was able to get past the Octavia/Amphora issue I was having last week. I got further and am now seeing timeouts connecting to the amphora. I replied to an email I got over the weekend but FYI here is what I am seeing. Would you have advice, ideas, or direction for this? I really appreciate all the help. | 12:53 |
| mephmanx | johnsom https://pastebin.com/bU8RF83N | 12:54 |
| mephmanx | if anyone else has an idea on what would cause the octavia/amphora issue I posted in the pastebin earlier, would appreciate all help. Thanks! | 13:27 |
| gthiemonge | mephmanx: I'm thinking about 2 reasons: 1) nova doesn't start the VM correctly 2) or the octavia management network is not setup properly (openstack console log show might display some useful info there) | 13:35 |
| mephmanx | the amphora vm seems to start ok...I can see a login prompt on the console screen... Dosent seem like there are errors in the logs there. I see networks and such. Im thinking the octavia management network has an issue... I use kolla to do the install so I know it is doing some/most of the setup. Im looking through there and comparing that to what is needed manually... | 13:37 |
| gthiemonge | I'm not familiar with kolla, but this point in important: https://docs.openstack.org/kolla-ansible/latest/reference/networking/octavia.html#network-and-subnet | 13:38 |
| mephmanx | That looks like the process I am reviewing. Ill print what I have soon, have a meeting to go to. Would you have an idea if I got you access to my env? | 13:53 |
| mephmanx | Here are the rules that I have now: | 14:45 |
| mephmanx | https://pastebin.com/xA7fyKFj | 14:45 |
| gthiemonge | mephmanx: you shound have a linux interface on your controllers named 'o-hm0', the octavia services use this interface to communicate with the amphorae | 14:50 |
| gthiemonge | https://github.com/openstack/kolla-ansible/blob/295c69b5eeb80b7d661d0f936dcede18c33d15c7/ansible/group_vars/all.yml#L243 | 14:51 |
| mephmanx | mine would just the the api_interface. https://github.com/mephmanx/openstack-scripts/blob/e3c6f3711fa13f0bb6a5fa7f11bad535aeefd7e3/globals.yml#L137 | 14:52 |
| gthiemonge | mephmanx: for instance, in a devstack env, I have: https://paste.opendev.org/show/807962/ | 14:53 |
| gthiemonge | basically the addresses of this subnet should be reachable from the controllers | 14:55 |
| mephmanx | hmmm...ok. Here is what I have... https://pastebin.com/ULAxR2YJ | 14:57 |
| mephmanx | hmm.... ok, so maybe I should have the loadbalancer manager subnet on the same subnet as the controllers since I dont have a separate interface for it? | 14:59 |
| gthiemonge | hm I've never deployed without a seperate interface :/ | 15:00 |
| mephmanx | interesting... ok, so you are saying I need to add another interface to the controllers that is connected to that network? All of my networks are virtual....Im not 100% sure if I can do that.... That is somewhere to look though...is it only the controllers that need to get there or does compute need it as well? | 15:02 |
| mephmanx | Im using a full install, not devstact. Take a look at my git repo for more details. | 15:03 |
| johnsom | mephmanx Hi. Just to catch up, you can boot an amphora instance, the console log shows it got an IP on the 10.1.0.0/24 lb-mgmt-net subnet, and you are getting connection timeouts in the worker log. | 15:22 |
| mephmanx | yes | 15:26 |
| mephmanx | johnsom correct | 15:26 |
| johnsom | Yeah, so it probably is an issue with the controllers not having access to the network. I know the kolla team has worked on that over the last year. They have an IRC channel too called #openstack-kolla and might have better answers than we will. | 15:28 |
| mephmanx | That makes sense... I do see a change in the kolla source about not having the mgmt subnet not on the same network as the default created network. I wonder if that is for security or other errors. If its security, then Im not worried about it, my setup will be fine. if its something else...I could potentially waste a lot of time with it. I could add another network to my setup and just connect them all that way.... | 15:30 |
| johnsom | There are a lot of different ways to wire it up. Fundamentally it is just a neutron managed network that the VMS get booted with. The trick is connecting the worker, health manager, and housekeeping processes to that network. The only concern about sharing is making sure you have dhcp with enough addresses, etc. | 15:33 |
| mephmanx | ok, so I could set it to the 10.0.200.0/24 assuming there is enough addresses? Thats good, that will make it easier... | 15:38 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!