2024-10-30T03:10:22.909Z

<jcollin> I couldn't reproduce this issue mentioned in the tracker. Please read the explanation <https://tracker.ceph.com/issues/65157#note-2>. If that's not the case, there was no enough information in the tracker. The tracker doesn't give much information on what went wrong or the logs that reveals the issue. @Milind Changire Please check and provide more information about the issue, so that I could work on it.

2024-10-30T12:37:33.949Z

<Igor Golikov> Sorry I missed daily. Spent time on debugging my PR, exploring fscrypt ( implemented new test there ) and started to explore  fscrypt tasks to take pick some

2024-10-30T12:48:24.925Z

<Venky Shankar> @Markuze This is the doc for dev workflow: <https://docs.ceph.com/en/latest/dev/developer_guide/basic-workflow/>

2024-10-30T12:48:42.650Z

<Venky Shankar> No worries, thx for the update @Igor Golikov

2024-10-30T13:37:43.839Z

<Markuze> Hey, folks.
There are two docs added in this pull request if there is interest in tracing or bootstrapping a kernel dev env.

2024-10-30T13:37:46.523Z

<Markuze> <https://github.com/ceph/ceph/pull/60561>

2024-10-30T19:04:17.612Z

<Igor Golikov> virtualenv/bin/teuthology-suite

2024-10-30T19:04:24.301Z

<Igor Golikov> Ahh now I see the ceph-object-corps issues. I guess this is a bunch of serialized objects, for backward compatibility testing. So now when I changed the MClientReply message, the unit test (I run all unit tests by the book) fails while deserializing this message from the corpus.... I found this page <https://docs.ceph.com/en/reef/dev/corpus/?highlight=corpus> and I understand that its normal not to decode old objects. In general, I assume that all cluster nodes must run the same Ceph version?

2024-10-30T19:17:04.108Z

<Igor Golikov> Hey folks, continuing to fight with teuthology 🙂
I run this command
`teuthology-suite --machine-type smithi --email [igolikov@ibm.com](mailto:igolikov@ibm.com) -p 100 --suite fs -k testing --force-priority --subset $((RANDOM % 1024))/1024 -d centos -c igolikov-wip-bug-64611` and I have completed build for centos in Shaman (<https://shaman.ceph.com/builds/ceph/igolikov-wip-bug-64611/45bce6a4e60aa15fb5b2479e834094c7ca56f0b1/default/414236/>)
Teuthology fails with `Packages for os_type 'ubuntu', flavor default and ceph hash '45bce6a4e60aa15fb5b2479e834094c7ca56f0b1' not found`
it looks like they just ignore the `-d centos` option

2024-10-30T19:25:22.858Z

<gregsfortytwo> @Patrick Donnelly has been doing some work on similar things so may have insight here.
Unfortunately the distro and kernel options are largely defunct right now, because we have tests that only work on one system — so they are specified as part of the suite yaml definitions

2024-10-30T19:26:06.575Z

<gregsfortytwo> I am not sure if the teuthology scheduler checks based on what is actually constructed, or just assumes you need all diaries

2024-10-30T19:27:07.626Z

<gregsfortytwo> If the former, you can edit the suite you are using to remove Ubuntu, and specify a suite repository/branch to use

2024-10-30T19:27:35.784Z

<Casey Bodley> could try `--filter centos` instead of `-d centos`

2024-10-30T20:00:11.964Z

<Patrick Donnelly> ^

2024-10-30T20:00:25.925Z

<Patrick Donnelly> -d just sets the "base os" but is generally overridden everywhere

2024-10-30T20:00:30.197Z

<Patrick Donnelly> s/base/default/

2024-10-30T20:03:05.931Z

<Patrick Donnelly> `(source teuthology/virtualenv/bin/activate; f() { teuthology-suite --machine-type smithi --email [pdonnell@redhat.com](mailto:pdonnell@redhat.com) --force-priority -p 95 --suite fs -k testing --subset $((RANDOM % 1024))/1024 --filter centos --ceph igolikov-wip-bug-64611 --dry-run  ; } ; f ) |& tee READ`

2024-10-30T20:03:07.574Z

<Patrick Donnelly> works for me

2024-10-30T20:03:19.482Z

<Patrick Donnelly> (note the --dry-run part)

2024-10-30T20:12:34.131Z

<Igor Golikov> I am already deep into debugging teuthology and I found that after merging configurations (I have no idea why do they need to merge singe configuration) it changes the OS to ubuntu. @Patrick Donnelly thanks for that, will try!

2024-10-30T20:24:50.696Z

<Igor Golikov> Another productivity issue - why do we need to trigger ALL configurations builds in Shaman? I would love to have an option to use something like tagging my branch in ceph-ci with say ‘build-config’ label, e g centos-x86 , and in this case they need to run only one build job instead of 7… for running teuthology suites we dont need all build configs available in Shaman

2024-10-30T20:26:22.921Z

<Igor Golikov> This is how I spent 3 hours today fighting with -d… Shaman has completed only centos build for my branch. So instead of waiting for ubuntu branch to be completed, I decided to run teuthology with -d … 

2024-10-30T21:48:22.240Z

<gregsfortytwo> Are the builders that far behind, or are there build failures on the other distros?

2024-10-30T23:03:47.734Z

<Patrick Donnelly> some history is documented in <https://github.com/ceph/teuthology/blob/main/docs/fragment_merging.rst> (but it's mostly covering a new scripting feature)