| opendevreview | OpenStack Proposal Bot proposed openstack/requirements master: Updated from generate-constraints https://review.opendev.org/c/openstack/requirements/+/821795 | 02:48 |
|---|---|---|
| *** johnsom_ is now known as johnsom | 15:00 | |
| *** hindret_ is now known as hindret | 15:00 | |
| opendevreview | OpenStack Proposal Bot proposed openstack/requirements master: update constraint for neutron-lib to new release 2.18.0 https://review.opendev.org/c/openstack/requirements/+/821859 | 15:49 |
| gmann | prometheanfire: I just notice requirement conflict which seems like no-conflict to me, I am seeing it 2nd time - https://zuul.opendev.org/t/openstack/build/951694d253164950979558f6bd3ee3ee | 18:26 |
| gmann | ERROR: Cannot install keystone==20.1.0.dev19 because these package versions have conflicting dependencies. | 18:26 |
| gmann | The conflict is caused by: | 18:26 |
| gmann | keystone 20.1.0.dev19 depends on PyJWT>=1.6.1 | 18:26 |
| gmann | The user requested (constraint) pyjwt===2.3.0 | 18:26 |
| gmann | 2.3.0 is >=1.6.1 | 18:26 |
| prometheanfire | gmann: maybe it's caps vs no caps? | 18:33 |
| gmann | prometheanfire: humm, I see cap everywhere except in zuul one https://codesearch.opendev.org/?q=pyjwt&i=nope&literal=nope&files=&excludeFiles=&repos= | 18:34 |
| gmann | and from there is picking no-cap one? | 18:35 |
| gmann | it does not fail always. I saw 2 times in this week and recheck pass | 18:35 |
| prometheanfire | maybe? https://review.opendev.org/821388 has someone wanting to switch to prettytable (lower case) | 18:35 |
| prometheanfire | intermittancy sucks even more | 18:36 |
| gmann | yeah replace is more work and block project side | 18:38 |
| prometheanfire | yep, I'm not certian it's necessary | 18:38 |
| prometheanfire | https://pypi.org/project/SeTuPtOoLs/ seems to redirect | 18:39 |
| gmann | but for pyjwt we can replace it to PyJWT in zuul which should not break any | 18:43 |
| gmann | fungi: ^^ | 18:43 |
| fungi | it's the same old pypi cdn issue we see from time to timw | 18:44 |
| opendevreview | Merged openstack/requirements master: update constraint for neutron-lib to new release 2.18.0 https://review.opendev.org/c/openstack/requirements/+/821859 | 18:46 |
| fungi | something happens in fastly (the cdn used by pypi) which causes some of its endpoints to be unable to reach the primary pypi backend, so they pull data from a fallback mirror the pypi devs maintain. unfortunately that mirror tends to frequently lag behind actual pypi contents and serves up stale indices which don't list things that made it into our constraints list, so jobs end | 18:46 |
| fungi | up failing because they're asked to use constraints which refer to what they see as nonexistent versions of some dependencies | 18:46 |
| fungi | this is made worse by the fact that (as we've discovered) the fallback mirror serves up a cache validity of 24 hours instead of 10 minutes like the real pypi uses, so our caching proxies hold onto those stale indices for far longer, causing them to accumulate and persist well after the problem clears up on pypi/fastly's end | 18:48 |
| gmann | fungi: ohk. I thought pyjwt (no cap ) is coming bcz zuul |registry requirements.txt having in no-cap | 18:49 |
| fungi | nah, entirely because the job asks pip to install pyjwt===2.3.0 and pip doesn't find any 2.3.0 in the pyjwt index it receives | 18:51 |
| fungi | either because that index is very outated, or sometimes truncated/empty | 18:51 |
| fungi | depending on how the problem manifests | 18:51 |
| fungi | has it happened again in recent minutes, or were these failures from earlier? | 18:52 |
| gmann | i see. I have seen it this week only, one in temest patch and one in cinder today | 18:53 |
| gmann | let's see if it become more frequent | 18:53 |
| fungi | well, we've been sending purge calls to fastly asking them to clear the index pages for failures people have mentioned to us, and also wiping our caches to make sure we don't continue to serve those to new builds | 18:54 |
| fungi | last cleanup was performed just prior to 07:00 utc today | 18:55 |
| fungi | so if you see similar failures for builds started after that time, then we probably need to repeat | 18:56 |
| gmann | ohk, today i saw is ~19 hr ago whihc is before 7 UTC. i will ping if I see again | 18:57 |
| gmann | thanks for information | 18:57 |
| fungi | yw | 18:57 |
| opendevreview | Emilien Macchi proposed openstack/requirements stable/ussuri: [stable-only] pin virtualenv https://review.opendev.org/c/openstack/requirements/+/821878 | 21:03 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!