| opendevreview | Dmitriy Rabotyagov proposed openstack/octavia stable/yoga: Fix barbican client with application credentials/trusts https://review.opendev.org/c/openstack/octavia/+/859913 | 08:03 |
|---|---|---|
| noonedeadpunk | Hi folks. Any chance to get these backports reviewed? https://review.opendev.org/c/openstack/octavia/+/859793 | 08:04 |
| gthiemonge | noonedeadpunk: sure, I'm going to take a look | 08:14 |
| opendevreview | Merged openstack/octavia stable/yoga: Fix prometheus-proxy service name in Red Hat-based distros https://review.opendev.org/c/openstack/octavia/+/862231 | 08:17 |
| noonedeadpunk | thanks! | 08:17 |
| gthiemonge | noonedeadpunk: do you think other stable branches are also affected? (I'm thinking about cherry-picking it to wallaby too) | 08:18 |
| opendevreview | Tom Weininger proposed openstack/octavia-tempest-plugin master: Add scenario tests for failover circuit breaker https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/837878 | 08:22 |
| noonedeadpunk | yup, pretty sure they are, but IIRC there was no clean backport | 08:22 |
| noonedeadpunk | so I did only all clean cherry-picks first to see if these are approriate at least before checking rebase | 08:23 |
| gthiemonge | noonedeadpunk: ok, I will backport it to wallaby (if it is doable), thanks! | 08:27 |
| opendevreview | Merged openstack/octavia stable/zed: Fix barbican client with application credentials/trusts https://review.opendev.org/c/openstack/octavia/+/859912 | 10:23 |
| opendevreview | Merged openstack/octavia stable/xena: Fix barbican client with application credentials/trusts https://review.opendev.org/c/openstack/octavia/+/859793 | 10:30 |
| opendevreview | Tom Weininger proposed openstack/octavia master: Remove support for split listener configuration https://review.opendev.org/c/openstack/octavia/+/862599 | 11:41 |
| opendevreview | Merged openstack/octavia stable/yoga: Fix barbican client with application credentials/trusts https://review.opendev.org/c/openstack/octavia/+/859913 | 11:46 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia stable/wallaby: Fix barbican client with application credentials/trusts https://review.opendev.org/c/openstack/octavia/+/862600 | 11:50 |
| opendevreview | Gregory Thiemonge proposed openstack/octavia master: Fix AllocateVIP.revert in amphorav2 https://review.opendev.org/c/openstack/octavia/+/862617 | 14:25 |
| opendevreview | Tom Weininger proposed openstack/octavia master: Remove support for split listener configuration https://review.opendev.org/c/openstack/octavia/+/862599 | 14:37 |
| opendevreview | Tom Weininger proposed openstack/octavia master: Remove support for split listener configuration https://review.opendev.org/c/openstack/octavia/+/862599 | 14:39 |
| opendevreview | Tom Weininger proposed openstack/octavia master: Remove support for split listener configuration https://review.opendev.org/c/openstack/octavia/+/862599 | 14:57 |
| noonedeadpunk | I was actually wondering lately - any reason not to have active_active topology for amphora? Ie - not much sense in it? | 16:57 |
| noonedeadpunk | As I can recall some blueprint, but can't recall reason why it was never done | 16:57 |
| noonedeadpunk | (or it was just some discussion) | 16:58 |
| gthiemonge | noonedeadpunk: we mentioned active_active with bgp during the PTG, there are some old specs in the repo | 17:02 |
| noonedeadpunk | well, it can be done even without bgp though? | 17:03 |
| noonedeadpunk | just matter of another keepalived instance and DNS RR | 17:03 |
| noonedeadpunk | with jinja should be easy to template that... But yes, 2 VIPs would be needed | 17:03 |
| noonedeadpunk | and that might lead a lot of complications with it... | 17:04 |
| noonedeadpunk | yep, actually PTG discussion pushed me thinking if bgp is required for that | 17:04 |
| noonedeadpunk | though it returned me back to if that is even needed. Like you said - amphora not designed for really high throughput. And current performance might be validly enough | 17:08 |
| noonedeadpunk | I see bunch of architectures for that.. None consider DNS RR for some reason though... | 17:11 |
| noonedeadpunk | just out if interest spawned now 2 haproxies behind keeplived with osa roles - and it works perfectly with DNS RR... And in case second fails - VIP is moved to first keepalived instance so DNS RR is not broken until it recovers | 17:16 |
| noonedeadpunk | so matter of allocating another VIP basically and adjusting keepalived tempalate... | 17:16 |
| johnsom | noonedeadpunk So, active-active has stalled a few times, mostly due to not having people to work on it. | 20:53 |
| johnsom | There are a bunch of details that surface once you start getting into it. | 20:53 |
| johnsom | Like, DNS RR, which can have issues with session persistence, etc. not to mention a DNS layer on top, health monitoring for the amps, etc. That is/was project "Kosmos" https://wiki.openstack.org/wiki/Kosmos | 20:55 |
| johnsom | There are also issues with session/state sync, growing/shrinking the pool (consistent hashing), etc... | 20:55 |
| johnsom | You can get a pretty high throughput with the amps, so yes, it's also a question of if there is a need enough for people to be able to finish developing it. | 20:57 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!