Friday, 2025-10-03

*** mtreinish_ is now known as mtreinish00:22
opendevreviewOpenStack Proposal Bot proposed openstack/project-config master: Normalize projects.yaml  https://review.opendev.org/c/openstack/project-config/+/96255702:12
hemanthHey I have the following patch in sunbeam-charms project https://review.opendev.org/c/openstack/sunbeam-charms/+/959080/10 where zuul shows Change has successfully merged but i do not see the commit in https://opendev.org/openstack/sunbeam-charms/commits/branch/main03:14
hemanthThe promote pipeline is failed and it is expected to fail. Does this have any affect on the merge 03:17
tonybhemanth: that obviously seems strange.   I'm having lunch ATM but if no-one else looks at it I will when I'm at my desk03:17
hemanththank you03:17
tonybwe are having issues with review.o.o ATM but I don't think they'd cause this03:17
tonybhow long did that change merge? just now or ages ago? where "ages ago" is more than an hour 03:18
hemanthcouple of hours ago03:19
hemanthand this one an hour ago https://review.opendev.org/c/openstack/sunbeam-charms/+/96289403:19
tonybokay.03:20
hemanthtonyb: I see the commits now in opendev.org/openstack/sunbeam-charms, thanks!04:45
tonybhemanth: I was just starting to look at it.  :/04:45
tonybhemanth: I don't think that the charm will get published to charmhub without another change merging04:51
hemanthtonyb: yeah I am fine with that.. I will handle it04:52
tonybokay04:52
opendevreviewIvan Anfimov proposed openstack/project-config master: wip  https://review.opendev.org/c/openstack/project-config/+/96292512:04
fungitonyb: hemanth: the cogent connectivity issues to review might also be impacting git replication to gitea13:25
fungionce thing settle down it would probably be a good idea to force re-replicate all projects just to make sure there aren't other commits missing in gitea too13:26
clarkbhemanth: tonyb: fungi triggered replication and I see that commit and a few others in gitea now. It was almost certainly the cogent outage as the timing lines up perfectly15:25
dansmithseems like network access into opendev servers is really slow today?15:31
dansmithgerrit and even the docs site take a long time to respond15:31
dansmithzuul seemed pretty responsive though15:31
clarkbdansmith: the docs site and gerrit are completely independent and in different cloud providers in different parts of north america. Gerrit was impacted by a large cogent outage last night that we've had to rerun background tasks for to sync things back up with the gitea servers. I suspect that is why you noticed slowness there. It should be back to normal at this point15:33
clarkbthough.15:33
clarkbdansmith: then the docs server is overwhelmed by clients doing slow reads filling all the slots. We've got a change in flight to increase the limits in apache15:33
dansmithmy devstack just timed out cloning from it15:33
fungicloning from gerrit (review.opendev.org) or gitea (opendev.org)?15:34
clarkband just == last 20 minutes or literally right now?15:34
dansmithreview.opendev and yes just now15:36
clarkbok ideally no one is cloning from review.opendev (you should use the giteas for that). I was just able to clone opendev/system-config and it is slower than I anticipate Receiving objects: 100% (135717/135717), 48.49 MiB | 871.00 KiB/s, done. But functional15:37
clarkblet me test from my ovh node to see if this could be slowness in cogent still15:37
dansmith(brb)15:37
clarkbI think it is cogent still being sad I get Receiving objects: 100% (135717/135717), 48.49 MiB | 10.28 MiB/s, done. from my ovh node that was not affected by the cogent issue15:38
dansmithclarkb: for unmerged patches? I figured cloning from gerrit was the only way15:38
dansmithare all those refs mirrored to gitea?15:38
clarkbdansmith: they are15:39
fungithey're all mirrored actually, yes15:39
clarkbbut also you don't typically clone to get unmerged patches15:39
clarkbbut either way the data should be in gitea15:39
clarkbfungi manually deployed the apache tunables update to the server hosting docs and now we shall see if this is sufficient to get ahead of the slow clients15:41
clarkbmtr shows a path that looks similar to what had problems last night. Maybe they had to reduce capacity to remove broken hardware from the path ro somethign?15:45
clarkbtheir status page only shows an ongoign issue with att in the atlanta region though15:46
clarkb(however last night I couldn't even reach their status page)15:46
sfernandhey folks! Regarding the failing NFS jobs with POST_FAILURE, adding the depends-on for the patch has helped us get past the immediate failure, so now we have access to the logs. Thanks a lot clarkb for the help! :)15:51
sfernandIt seems the underlying issue is that qemu-img convert is randomly getting stuck when it tries to write to the NFS mount. When it stalls, I see messages in the syslog like "task nfsd:71407 blocked for more than 122 seconds" [1] right after the qemu-img call. 15:51
sfernandIs there any chance we could get a node held so I can ssh in and investigate further? I suspect it's a resource issue on the node itself (like memory exhaustion or I/O limits)15:51
sfernand[1] https://zuul.opendev.org/t/openstack/build/8e77fc30cdfe4ac98f2d4ef663c57f20/log/controller/logs/syslog.txt#442915:51
clarkbsfernand: yes, if you let us know the change number and job you wish to hold I can put that in plce for you15:51
clarkbsfernand: basically I tell zuul to hold things then you recheck to trigger the failure and that will preserve the test env15:52
sfernandawesome! change is https://review.opendev.org/c/openstack/devstack-plugin-nfs/+/952476 15:58
sfernandjob is devstack-plugin-nfs-tempest-full, we may need to retrigger a couple time to see it fails15:58
clarkbsfernand: the hold request is in place. recheck away. Once you've got your failure let us know what your ssh pubkey is and we can add it to the environment16:01
sfernandwill do. thanks!16:07
dansmithclarkb: fungi sorry, was on a call16:34
dansmithwe clone to get unmerged patches all the time when we're testing things before merge... I'm not sure why that's unusual16:35
dansmithso you're saying I can use git.opendev but with the same refs/ path that gerrit gives me for a given patch and revision?16:35
dansmithsurely that mirroring can't be instant, right? so how long after someone pushes a rev to gerrit can I see it from the gitea side?16:36
fungidansmith: not instant, but usually sub-second17:13
fungibasically every time a change updates, gerrit internally triggers replication to gitea and those tasks can queue but usually get processed near-instantly17:14
fungiand yeah, the refs/... path should be identical17:15
fungithey're not browseable because gitea lacks the ability to browse named refs, but you should still be able to git fetch them17:15
clarkbdansmith: zuul only fetches the new changes from gerrit. It doesn't do a new clone each time17:22
clarkbwe try as much as possible to get peopel to not clone from gerrit itself17:23
clarkbthat is why we have a farm of replicas17:23
clarkband yes usually subsecond but for larger repos like nova it will be longer. However still measurable in a few seconds I think17:23
clarkbbut I don't think that is why you're seeing problems. I think the issue is related to the cogent problems from last night. Now we don't have packet loss but do appear to have much reduced bandwidth17:24
dansmithokay, I'll try to start cloning from gitea then17:31
dansmithclarkb: I  understand zuul doesn't, but for a fresh devstack in a VM, it's easier to let it clone than pre-provision the repos, especially since clean and unstack blow away /opt 17:31
fungidansmith: the other thing to keep in mind is that if you "clone" at a specific gerrit change, you're getting that change's parentage rather than how zuul is actually going to test it. instead you want to fetch and cherry-pick any involved changes onto the target branch tip17:33
dansmithfungi: no, I really don't.. what I want is to test a given patch set the way it is in review.. I'm not trying to reproduce what zuul is going to do when/if the merge happens17:34
fungiwe avoid gratuitously rebasing chages in review, which can lead to them having an outdated commit history compared to the branch state unless there's an actual merge conflict, but yeah if your goal is not to replicate what zuul is going to test then i guess that's sufficient17:34
sfernandclarkb: job failed on first run20:30
sfernand https://zuul.opendev.org/t/openstack/build/339e847cb5de4ae0aa5a5dc8705da8c820:30
fungisfernand: looks like your held node is listed near the top of https://zuul.opendev.org/t/openstack/nodes so that worked. what ssh private key do you want granted access to it?20:36
sfernandwhat is the preferable way to share a public key with you20:38
fungiyou can dump it here in irc or /msg it to me, or if you have it up at a url in, like, launchpad or github i can pull it from there just let me know the url20:38
fungiwhatever works for you really20:40
fungier, share the public key not the private key, of courser20:40
fungithe file with the .pub extension, e.g. ~/.ssh/id_ed25519.pub20:41
fungii said private initially, totally not what i meant20:41
sfernanddont worry I got it, just sent you20:43
fungiyep, you should be all set, let me know if not20:43

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!