ceph - sepia - 2024-07-02

Timestamp (UTC)Message
2024-07-02T04:05:01.095Z
<Suyash Dongre> Hi! I was able to create a pipeline in Jenkins, the clang-tidy checks that I wanted to run are working as expected. Now I am having a problem with polling. Jenkins is polling but even if there are changes (new commits) it is not able to start the build because no changes were found according to it.

This is how I have set up the configuration:

I have set the GitHub project to <https://github.com/ceph/ceph/>

I have ticked the poll SCM option and this is the schedule: `H/5 * * * *`

Is something missing? what can I do to fix this?

Thanks!
2024-07-02T04:08:30.013Z
<Suyash Dongre> Hi! I was able to create a pipeline in Jenkins, the clang-tidy checks that I wanted to run are working as expected. Now I am having a problem with polling. Jenkins is polling but even if there are changes (new commits) it is not able to start the build because no changes were found according to it.

This is how I have set up the configuration:

I have set the GitHub project to <https://github.com/ceph/ceph/>

I have ticked the poll SCM option and this is the schedule: `H/5 * * * *`

Am I missing something? what can I do to fix this?

Thanks!
2024-07-02T06:53:01.891Z
<Adam Kraitman> I focused on migrating builders to jammy,centos9 & arm64 were most of the load is, there is only one job using this label - vagrant&&libvirt&&centos9 ?: https://files.slack.com/files-pri/T1HG3J90S-F07AQFGQC7N/download/image.png
2024-07-02T09:02:34.254Z
<Rongqi Sun> Hi, would it be possible that jobs of such as 'ceph-dev-new-setup' and 'ceph-pull-requests-arm64' are running at the same time? Seems like from arm64 servers' view, when it is running make check jobs, the environment is not 'clean' at all which means e.g. I could see 'building rpm' process.
2024-07-02T10:18:30.007Z
<Dan Mick> It's possible, but the environment should be pretty separate.  For rpm builds there are some global side effects...things like installed repos and rpms.  We've never gotten around to adding brew/koji-type isolation to the upstream builds.  Hopefully we can get at least container isolation soon.
2024-07-02T10:20:29.658Z
<Dan Mick> Only guessing, but I think there's a legacy job config setting that will build regardless of whether there's a new commit or not.  I assume there's something similar for pipelines.
2024-07-02T10:24:30.763Z
<Dan Mick> So I discovered yesterday that the [quay.ceph.io](http://quay.ceph.io) pruner process was broken and attempted to fix it.  It's a bit fragile and depends on assumptions about tag names and numbers of containers per sha1 that were modified several times in the last few years.  I fixed some of them, but that exposed a problem where it would falsely consider some "sha1"-style tags to be orphaned and deleted them.  The result is that there are a few tens of images that are still present but are missing their sha1 tags, which is what is used from teuthology jobs to schedule them.
2024-07-02T10:26:24.481Z
<Dan Mick> I have code to identify the missing tags, and I'm close to code that will automate re-adding them, but it's not quite working yet, and I can't stay awake any longer.  I will either get that code working or add the missing tags manually tomorrow.
2024-07-02T10:26:58.820Z
<Dan Mick> Sorry for the inconvenience.  It's a more-fragile process than I'd like.
2024-07-02T14:06:30.450Z
<Guillaume Abrioux> no @Adam Kraitman
2024-07-02T14:07:32.125Z
<Guillaume Abrioux> please, can you add back more workers ?
2024-07-02T14:08:07.649Z
<Guillaume Abrioux> there used to be at least 3 workers here
2024-07-02T14:59:49.426Z
<Adam Kraitman> Yeah I will add two more
2024-07-02T15:01:17.460Z
<Guillaume Abrioux> thanks!
2024-07-02T15:01:36.175Z
<Guillaume Abrioux> it's gonna block the 19.1.0 release testing i believe
2024-07-02T15:51:23.466Z
<Adam Kraitman> I don't see anything waiting in the queue in the past 10 days: https://files.slack.com/files-pri/T1HG3J90S-F07BDNBUASC/download/image.png
2024-07-02T17:01:08.032Z
<Guillaume Abrioux> that doesn't mean it doesn't have to be addressed
2024-07-02T17:01:31.932Z
<Guillaume Abrioux> workers are broken since more than 10 days
2024-07-02T17:01:44Z
<Guillaume Abrioux> because of the c8 > c9 migration
2024-07-02T17:02:23.560Z
<Guillaume Abrioux> why would I schedule a run if I know it's gonna fail or worse, queue for ever
2024-07-02T17:02:37.474Z
<Guillaume Abrioux> ceph-volume tests trigger more than 10 jobs
2024-07-02T17:02:59.091Z
<Guillaume Abrioux> if there's only one worker, each time we raise a ceph-volume PR, jobs will take hours and hours to complete
2024-07-02T17:03:16.673Z
<Guillaume Abrioux> there are always been at least 3 workers
2024-07-02T17:04:54.203Z
<Guillaume Abrioux> which is already too few
2024-07-02T17:05:32.236Z
<Guillaume Abrioux> can't we migrate everything from [2.jenkins.ceph.com](http://2.jenkins.ceph.com) to [jenkins.ceph.com](http://jenkins.ceph.com) by the way?
2024-07-02T17:05:33.579Z
<Adam Kraitman> I see in that case I will add more then two
2024-07-02T17:06:05.488Z
<Guillaume Abrioux> if I remember well, [2.jenkins.ceph.com](http://2.jenkins.ceph.com) was mostly for ceph-ansible
2024-07-02T17:06:26.551Z
<Guillaume Abrioux> ceph-ansible is in maintenance mode, not to say deprecated
2024-07-02T17:06:27.645Z
<Adam Kraitman> Yes I can migrate all the builders from 2.jenkins
2024-07-02T17:07:05.245Z
<Guillaume Abrioux> so it means we can take resources from [2.jenkins.ceph.com](http://2.jenkins.ceph.com), I think
2024-07-02T17:08:34.262Z
<Guillaume Abrioux> I think we should have more collaboration/coordination when it comes to allocating/moving Jenkins resources
2024-07-02T17:08:59.556Z
<Guillaume Abrioux> because if you'd ask about [2.jenkins.ceph.com](http://2.jenkins.ceph.com)
2024-07-02T17:09:06.443Z
<Adam Kraitman> I already migrated many builders from 2.jenkins to [jenkins.ceph.com](http://jenkins.ceph.com) , I can leave only few builders connected to 2.jenkins if we don't need more then that
2024-07-02T17:09:16.115Z
<Guillaume Abrioux> you would probably realize this jenkins instance is almost ... useless nowadays 🙂
2024-07-02T17:09:41.594Z
<Guillaume Abrioux> yeh but that's exactly the case
2024-07-02T17:09:58.573Z
<Guillaume Abrioux> we discover, by surprise, that resources are getting moved from an instance to another
2024-07-02T17:10:25.768Z
<Guillaume Abrioux> a notice would be appreciated 🙂
2024-07-02T17:11:31.512Z
<Guillaume Abrioux> and even for ceph-volume jobs
2024-07-02T17:11:48.790Z
<Guillaume Abrioux> they are not using [2.jenkins.ceph.com](http://2.jenkins.ceph.com)
2024-07-02T17:12:13.619Z
<Guillaume Abrioux> for more than 2 weeks i can't schedule jobs because it's broken
2024-07-02T17:12:23.858Z
<Guillaume Abrioux> so i'm patiently waiting it gets fixed
2024-07-02T17:12:46.264Z
<Guillaume Abrioux> and because it's broken, there's no load, because there's no load you state you can take away some resource for other projects
2024-07-02T17:12:55.080Z
<Adam Kraitman> Can you start when there are 3 builders or you need more to start ?
2024-07-02T17:13:05.392Z
<Guillaume Abrioux> once again, it sounds like a double punishment
2024-07-02T17:14:24.458Z
<Guillaume Abrioux> 3 builders are the very minimum I'd say, I can probably settle for that for now, but I might pester you in the future
2024-07-02T17:15:34.766Z
<Guillaume Abrioux> moving jenkins resources like that has impact on developpers, it should be done carefully
2024-07-02T17:16:05.891Z
<Adam Kraitman> The Centos8 EOL was a surprise to us so I had to migrate many builders
2024-07-02T17:16:16.546Z
<Guillaume Abrioux> yeh but migrate from c8 to c9 is a thing
2024-07-02T17:17:00.736Z
<Guillaume Abrioux> disallocating a resource from one project in favor of another project is another matter
2024-07-02T17:18:31.894Z
<Guillaume Abrioux> I'm sorry if I come across as rude, but it feels to me like I've been struggling with the CI for a while now 😕
2024-07-02T17:20:29.440Z
<Guillaume Abrioux> I don't want to stress you out more than that but I had to raise this concern
2024-07-02T17:20:48.907Z
<Guillaume Abrioux> I have to run, we can follow up on this tomorrow if you want
2024-07-02T17:21:35.789Z
<Guillaume Abrioux> good evening/night 🙂
2024-07-02T22:28:24.705Z
<Dan Mick> I think I've fixed up the tags.  There were a few more bugs to squash down.

Any issue? please create an issue here and use the infra label.