| opendevreview | chenker proposed openstack/governance master: Add RISC-V https://review.opendev.org/c/openstack/governance/+/988896 | 12:43 |
|---|---|---|
| TheJulia | gouthamr: are *we* (the community) really aware of folks still operating with Postgres 9 years later? | 13:42 |
| fungi | people pop up on the mailing list from time to time trying to adjust queries to work on pgsql, but don't typically provide much context for their use of it | 14:10 |
| fungi | i get the impression that the few who are still out there have accepted that it's up to them to make it work if it's really that important to them | 14:11 |
| TheJulia | I know there are shops which are fanatical towards one or the other. I just don't know how much sense it really makes and it feels like the discussion keeps coming back to "well do this" and we've never really had solid data beyond preference | 14:30 |
| opendevreview | Tatiana Ovchinnikova proposed openstack/governance master: Seed security liaison data from VMT wiki https://review.opendev.org/c/openstack/governance/+/996563 | 16:09 |
| gouthamr | TheJulia: I was trying to signal that tkajinam's request will be the final nail in the coffin.. projects dropping testing was one thing, removing the oslo.db integration will mean these operators can't use it any more | 16:52 |
| gouthamr | afaik we still run at least migration tests if not e2e tests with postgresql across lots of openstack services.. | 16:53 |
| gouthamr | these pretend to be "unit tests" | 16:54 |
| opendevreview | Tatiana Ovchinnikova proposed openstack/governance master: Seed security liaison data from VMT wiki https://review.opendev.org/c/openstack/governance/+/996563 | 17:37 |
| tkajinam | gouthamr TheJulia , I remember I saw some questions/problems related to PostgreSQL support several years back but I've not seen such for recent 2-3 years (IIRC). So I hope no one is using postgres any more | 18:02 |
| gouthamr | tkajinam: here's one that i saw: https://o3k.io/ | 18:03 |
| tkajinam | oh, that's interesting | 18:03 |
| gouthamr | tkajinam: otoh, hope the cursive blocker was resolved.. i poked some people downstream to get some attention from the barbican side.. ty for picking that up, and if you need any more help, lemme know | 18:04 |
| fungi | well, that's not openstack, it's a minimal compatible api vibe-coded in golang | 18:04 |
| tkajinam | my small concern is around the standalone use (for example I heard some people are using mistral as a standalone service though I don't know what is their db backend) | 18:04 |
| tkajinam | but I'm not much concerned about usage in actual "OpenStack" env since neutron already dropped its support | 18:04 |
| fungi | https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/message/NPOHAAPCMTZR6HPMLRVDQXPTFFQ2EVGG/ was the most recent mentiom of postgres i saw on openstack-discuss, roughly 2 months ago | 18:04 |
| tkajinam | (hoping that people are not stuck at old releases) | 18:04 |
| gouthamr | from that email: "After migrating all postgres data" - i wonder if that means they were moving off of postgres | 18:05 |
| fungi | "Re: using postgresql in production" https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/message/GIASU77VZQLVKD3KKJAYGSPFGTWGJPOJ/ was february | 18:05 |
| fungi | "...Recently I am experimenting using PG for a Victoria OpenStack..." | 18:06 |
| fungi | frightening, but yeah "stuck on old versions" seems to be a good guess | 18:06 |
| gouthamr | ouch | 18:06 |
| fungi | the start of that thread claims that postgres is very popular in china | 18:07 |
| tkajinam | hmm | 18:10 |
| tkajinam | because we are at non slurp cycle, we may need to delay the removal at least for 2 more cycles, but I want to gather some inputs and get the global conclusion so that we won't keep the support in current partial state | 18:10 |
| fungi | similarly "nova-manage fails when upgrading to Antelope" https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/6I2X74ATIEPKLLVRUDDWG4NQXWMY67Z2/#QQ7GUHDFY2IYC65O5YX2HXCHHX3CGBPH | 18:11 |
| fungi | in keeping with the old versions theory | 18:11 |
| tkajinam | gouthamr, regarding cursive, it's now almost unblocked. the CI fix has landed and I can now merge patches there | 18:11 |
| tkajinam | https://review.opendev.org/c/openstack/releases/+/993729 is still pending but may be merged soon. | 18:12 |
| tkajinam | I've merged a bunch of clean ups and modernization (packaging related). I will propose a few new releases. (actually I already did) | 18:13 |
| tkajinam | gouthamr, thanks for your help :-) | 18:13 |
| fungi | tkajinam: and i approved 993729 just now | 18:15 |
| gouthamr | neat! :) | 18:16 |
| tkajinam | fungi, thanks ! | 18:19 |
| Mike-- | fungi / gouthamr / TheJulia / tkajinam: to chime in from a "future" openstack operator. When we were orienting on which DB backend we started with pgsql as the kubernetes operator was more mature than mariadb/mysql. That gap seems to have closed a bit, but hey adding context | 18:20 |
| Mike-- | I remember talking in #openstack whether pgsql was supported or not, and the answer there was 'not really tested' so we moved away from pgsql | 18:21 |
| tkajinam | ah, that's interesting | 18:21 |
| tkajinam | in physical deployments I think mysql/mariadb provides much better experience for clustering (mainly thanks to galera) than postgresql. it sounds like the situation is a bit opposite in k8s deployments | 18:23 |
| Mike-- | we used to run galera clusters (3 AZ's multi-master) for years (2015 - 2022) so we are not afraid to use it, but k8s operators strongly preferred pgsql using the operator | 18:24 |
| gouthamr | this is good information, ty for sharing Mike-- - our hope then would be that someone offers to help with the postgres maintenance in oslo.db, and get services to test it.. stuff like what neutron did could hurt though, they did ask for help before they ripped out all the postgres compatibility etc from neutron itself.. | 18:43 |
| Mike-- | yes just adding context, we have since switched to the mariadb operator which has improved quite a bit | 18:45 |
| Mike-- | running an openstack cloud on something that is somewhat tested did not feel great ;D | 18:46 |
| gouthamr | yes! | 18:46 |
| gouthamr | i wonder where all we've said we support postgresql.. (or hidden the fact that it's not tested well, or that there's partial support)... https://docs.openstack.org/install-guide/environment-sql-database.html is one place | 18:48 |
| Mike-- | we started from the documentation which did not really make it clear and was interpretted as both are fully supported | 18:48 |
| gouthamr | ++ | 18:48 |
| gouthamr | someone mentioned that in one of the mail threads we discussed above | 18:49 |
| fungi | skimming https://codesearch.opendev.org/?q=postgres might turn up some examples | 18:50 |
| Mike-- | maybe add pgsql, postgres etc (the unique keywords) | 18:50 |
| gouthamr | ah! OSH deploys with postgres! | 18:50 |
| fungi | yes, just spotted that as well | 18:50 |
| Mike-- | without nudging either way, but if (1) pgsql is not well tested and (2) it is a pain to keep supporting it in oslo.db the logical route would be to mark it retired... my 0.02 | 18:51 |
| Mike-- | I believe genestack (rackspace) wraps OSH but uses the mariadb operator. And I know another commercial company bundles mysql with OSH | 18:52 |
| Mike-- | not too sure how widespread pgsql in OSH is used tbh | 18:53 |
| fungi | the search results for the openstack-manuals repo turn up a few hits claiming postgres is "supported" | 18:53 |
| gouthamr | yeah the aspect is the "distributed" nature of such a decision.. we know now that keystone became visibly incompatible circa 2023/Antelope, and neutron had to drop testing/code compatibility in 2024 (Dalmatian?), and ironic, manila, others dropped test jobs because they include all these components in their CI jobs.. so.. its playing out painfully slowly instead of a concerted effort | 18:53 |
| fungi | in the install guide specifically | 18:54 |
| Mike-- | the pgsql in rackerlabs is for their nautobot instance (FYI) | 18:55 |
| fungi | the security guide makes some postgres references too | 18:57 |
| fungi | aodh, barbican, blazar, cinder, cloudkitty, designate, glance, heat, keystone, magnum, manila, masakari, mistral, molteniron, nova, placement, rally, tacker, taskflow, tooz, trove, zun all include some vestiges of postgresql docs/support/testing | 19:07 |
| fungi | as do several neutron stadium repos still | 19:08 |
| fungi | and basically every upstream deployment project is still plumbed for postgresql as an option | 19:08 |
| fungi | given that, i can definitely see where someone might mistakenly come the the conclusion it's a fine choice | 19:09 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!