ceph - ceph-devel - 2024-06-20

Timestamp (UTC)Message
2024-06-20T03:20:37.942Z
<badone> @Casey Bodley would you have time to take another look at <https://tracker.ceph.com/issues/59380#note-48> since we have a lot more information now?
2024-06-20T09:59:01.023Z
<zigo> Hi there!
2024-06-20T09:59:01.530Z
<zigo> Is there a reason why my pull requests recieved no reviews?
2024-06-20T09:59:01.841Z
<zigo> https://github.com/ceph/ceph-container/pull/2231
2024-06-20T11:03:47.243Z
<Ilya Dryomov> @Patrick Donnelly Could it be you who is experimenting with tags on the tracker (colors, sidebar, etc)?
I noticed that automatic backport ticket creation broke sometime yesterday and remains broken today
Here in an example error:
``` INFO:root:Processing 865 issues with status Pending Backport
Examining issue#66438 (1/865)
Traceback (most recent call last):
  File "/home/runner/work/ceph/ceph/src/script/backport-create-issue", line 397, in <module>
    iterate_over_backports(redmine, issues, dry_run=args.dry_run)
  File "/home/runner/work/ceph/ceph/src/script/backport-create-issue", line 342, in iterate_over_backports
    mark_as_processed(r, issue)
  File "/home/runner/work/ceph/ceph/src/script/backport-create-issue", line 311, in mark_as_processed
    if tag_backport_processed not in field['value']:
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: argument of type 'NoneType' is not iterable
Examining issue#66420 (2/865)
Error: Process completed with exit code 1.```
2024-06-20T12:39:38.554Z
<frickler> failing assumptions about structured variables, such a classic *scnr* ;)
2024-06-20T13:14:07.054Z
<Matheus de Oliveira Morozetti> Hello Everyone I'm facing a problem with my storage my error is
```[WRN] TOO_MANY_PGS: too many PGs per OSD (361 > max 250)```
seems I have too much PG per OSD, my cluster is new and I don't have so much data. My total PG is 3137 and my total of pools is 26, I got some OSD with much PG

```	•	OSD 16: 450 PGs
	•	OSD 2: 417 PGs
	•	OSD 3: 418 PGs
	•	OSD 20: 416 PGs```
I don't know how PG are handlend with my OSD's, I'm using default value for PG per OSD's

```ceph config get mon mon_max_pg_per_osd
1500```
2024-06-20T14:01:18.579Z
<nehaojha> The Ceph User Dev Meeting is starting now <https://meet.jit.si/ceph-user-dev-monthly>!
2024-06-20T15:15:35.622Z
<Patrick Donnelly> No, I didn't do that.
2024-06-20T15:16:04.420Z
<Patrick Donnelly> and I think that it looks pretty annoying TBH
2024-06-20T15:17:30.455Z
<Ilya Dryomov> I don't disagree 🙂
2024-06-20T15:18:22.359Z
<Patrick Donnelly> I think @Leonid Usov requested tags for backport tickets which I did do
2024-06-20T15:18:33.543Z
<Patrick Donnelly> and maybe that broke the script (again) but I thought we also fixed it
2024-06-20T15:20:42.935Z
<Patrick Donnelly> the script works for me now, maybe try re-running on main?
2024-06-20T15:20:53.416Z
<Ilya Dryomov> AFAICS `backport-create-issue` hasn't changed in months
2024-06-20T15:21:30.470Z
<zigo> Casey Bodley: Is it still relevant for me to make a MR against the ceph-container repo? What has been decided?
2024-06-20T15:21:51.926Z
<Patrick Donnelly> oh, that
2024-06-20T15:21:58.644Z
<Patrick Donnelly> i thought we were talking about ceph-backport.sh
2024-06-20T15:22:25.476Z
<Patrick Donnelly> yes, it's broken
2024-06-20T15:22:26.050Z
<Patrick Donnelly> <https://github.com/ceph/ceph/actions/workflows/create-backport-trackers.yml>
2024-06-20T15:22:31.834Z
<Patrick Donnelly> I'll take a look
2024-06-20T15:23:37.867Z
<Casey Bodley> hey zigo. i don't know that any decision was made there. i see that @Ilya Dryomov tagged @Guillaume Abrioux for review on <https://github.com/ceph/ceph-container/pull/2231>
2024-06-20T15:24:01.045Z
<zigo> Ok, thanks.
2024-06-20T15:24:26.553Z
<Ilya Dryomov> Since the breakage has to do with the tag, I thought those visual changes were related
2024-06-20T15:24:46.522Z
<zigo> Casey Bodley: I also wonder if it would be somehow possible to have Debian images hosted at quay.io/ceph ?
2024-06-20T15:25:34.152Z
<Ilya Dryomov> The timing fits -- it broke yesterday and I first that colorful sidebar yesterday
2024-06-20T15:25:48.707Z
<Ilya Dryomov> The timing fits -- it broke yesterday and I first saw that colorful sidebar yesterday
2024-06-20T15:26:20.665Z
<Casey Bodley> that might be a question for Guillaume? i'm not well versed in this container business
2024-06-20T15:26:22.640Z
<Patrick Donnelly> last successful run: <https://github.com/ceph/ceph/actions/runs/9585983955>
2024-06-20T15:26:31.686Z
<Casey Bodley> ^ or @Adam King maybe?
2024-06-20T15:27:05.393Z
<Ilya Dryomov> Sorry, I did it just because it's a ceph-container PR
2024-06-20T15:27:29.196Z
<Patrick Donnelly> It looks like someone added a "redmineup_tags" plugin
2024-06-20T15:27:31.621Z
<Patrick Donnelly> cc @Adam Kraitman
2024-06-20T15:27:47.564Z
<Ilya Dryomov> I'm lacking context as far as any discussion that might have taken place
2024-06-20T15:28:23.726Z
<Adam King> It's still normally Guillaume that looks at ceph-container I think. I've only contributed small bits there.
2024-06-20T15:28:30.281Z
<Ilya Dryomov> Sorry, I did it just because I saw zigo's message and because it's a ceph-container PR
2024-06-20T15:28:52.775Z
<Casey Bodley> i believe the context comes from <https://github.com/ceph/ceph-container/issues/2171>
2024-06-20T15:31:23.451Z
<zigo> Yeah, exactly, I am wondering what's going to happen, and if I'm right to push a PR over there, or if I should do it eslewhere...
2024-06-20T15:32:53.591Z
<Matthews Jose> Hello Everyone,
I am trying to parse the hearbeat messages that OSD sends to the MONs within the cluster network. But going through the documentation I cannot find a packet headers for this message. Is there documentation on this, I am currectly look at MOSDBEACON.h, (from wireshark it is a 96bytes message). Could someone point me the right direction where I can find this header format with size of each field for the heart beat messages. Thank you.
2024-06-20T15:33:18.779Z
<Ilya Dryomov> In that PR, Ken suggested:
> There are a few people in the Ceph community who have discussed maintaining Ubuntu- or Debian-based container images. To provide a common place for these, I suggest renaming this GitHub repo from `ceph-container` to `ceph-containers` or `alt-containers`
2024-06-20T15:35:18.743Z
<Ilya Dryomov> ... so it seems like zigo's PR is in order (against the right repo, even if the repo hasn't been renamed and is still being used for CentOS-based images as well)
2024-06-20T15:35:52.221Z
<zigo> :)
2024-06-20T15:37:44.218Z
<zigo> I have to go, ttyl
2024-06-20T15:40:26.734Z
<Adam Kraitman> Hey @Patrick Donnelly I added that plugin we had this plugin on the old version but it's not supported in the current version so I had to add redmineup_tags instead
2024-06-20T15:48:42.521Z
<Ronen Friedman> An interesting example of a pessimizing 'final':
<https://x.com/DrawsMiguel/status/1801076304000737685>
(due to inability to use EBO in this instance).
2024-06-20T15:52:16.541Z
<Patrick Donnelly> it seems we have two "Tags" custom fields  now
2024-06-20T15:52:18.931Z
<Patrick Donnelly> I'll try to consolidate
2024-06-20T15:52:50.757Z
<Patrick Donnelly> maybe it was removed?
2024-06-20T15:52:53.999Z
<Patrick Donnelly> tag id 3 is gone
2024-06-20T15:52:59.479Z
<Patrick Donnelly> custom field id 3 is gone
2024-06-20T15:55:51.015Z
<Patrick Donnelly> If you edit a ticket like <https://tracker.ceph.com/issues/66420>
2024-06-20T15:55:55.831Z
<Patrick Donnelly> you'll see there are two "Tags" fields
2024-06-20T15:55:59.488Z
<Guillaume Abrioux> hum
2024-06-20T15:56:08.085Z
<Patrick Donnelly> one seems to be custom_field id 31
2024-06-20T15:56:16.670Z
<Patrick Donnelly> and the other, I don't know where it is from
2024-06-20T15:58:49.288Z
<Guillaume Abrioux> the debian based OS support doesnt receive any engineering effort for a while
2024-06-20T15:59:10.090Z
<Guillaume Abrioux> to be honest, I'm not sure what the current status of it is at the moment
2024-06-20T16:02:04.239Z
<Patrick Donnelly> @Ilya Dryomov I reconfigured the plugin to create a "cloud" which is more compact: <https://tracker.ceph.com/projects/cephfs/issues>
2024-06-20T16:02:06.623Z
<Patrick Donnelly> what do you think?
2024-06-20T16:02:10.527Z
<Guillaume Abrioux> zigo, I don't mind merging this patch given that it doesn't touch anything regarding CentOS builds if that can help you for the time being, one thing to keep in mind is that there's no tests at all, the CI has just no coverage for debian
2024-06-20T16:02:48.714Z
<Patrick Donnelly> @Adam Kraitman @Ilya Dryomov <https://github.com/ceph/ceph/pull/58174>
2024-06-20T16:02:48.803Z
<Adam Kraitman> One is from the plugin(where there is a drop down list) and the 2nd we had it from the past but if no one is using it I can remove it
2024-06-20T16:03:41.417Z
<Adam Kraitman> One is from the plugin(where there is a drop down list) and the 2nd Tags field we had it from the past but if no one is using it I can remove it
2024-06-20T16:04:46.987Z
<Casey Bodley> > I also wonder if it would be somehow possible to have Debian images hosted at [quay.io/ceph](http://quay.io/ceph)?
any idea what that would involve, @Guillaume Abrioux? or who to ask about that?
2024-06-20T16:05:22.910Z
<Guillaume Abrioux> we just need to make our job push images to [quay.io](http://quay.io)
2024-06-20T16:05:28.033Z
<Guillaume Abrioux> that shouldn't be a big deal
2024-06-20T16:09:37.677Z
<Guillaume Abrioux> out of ignorance and curiosity, I thought we wanted to stop supporting builds other than CentOS?
2024-06-20T16:12:12.509Z
<Casey Bodley> i don't think that's the case. <https://docs.ceph.com/en/latest/start/os-recommendations/#platforms> lists the distros that we currently support
2024-06-20T16:12:31.408Z
<Guillaume Abrioux> that's something I believe I've heard but perhaps nothing has never been clearly decided in the end
2024-06-20T16:13:18.546Z
<Guillaume Abrioux> yeh but this list isn't about the base OS for building ceph containerized images
2024-06-20T16:14:14.517Z
<Casey Bodley> afaik our upstream containers have only ever supported centos 8 (and recently 9)
2024-06-20T16:14:20.331Z
<Guillaume Abrioux> in a containerized mindset, you shouldn't care about the underlying OS in your container image
2024-06-20T16:15:41.997Z
<Guillaume Abrioux> I mean, personally I don't care, if for some reason someone feels like he needs a debian based image, that's perfectly fine with me 🙂
2024-06-20T16:15:46.880Z
<Casey Bodley> i agree that the ceph project should pick a container base distro and support just that. but for context, zigo maintains the ceph packages for the debian project. he's interested in providing debian-based containers
2024-06-20T16:17:11.159Z
<Guillaume Abrioux> I just remember some discussions - but nothing clearly stated - where it was considered to drop any builds other than CentOS
2024-06-20T16:17:50.254Z
<Guillaume Abrioux> by "any builds", I mean, any ceph container image builds
2024-06-20T16:19:27.260Z
<Guillaume Abrioux> I left a comment in the PR, I'll take a look for pushing the images to [quay.io](http://quay.io)
2024-06-20T16:21:03.377Z
<Adam Kraitman> Yeah "cloud" is more compact I also think it's better
2024-06-20T16:24:38.901Z
<Guillaume Abrioux> I mean, personally I don't mind if for some reason someone feels like he needs a debian based image, that's perfectly fine with me 🙂
2024-06-20T16:27:54.395Z
<Patrick Donnelly> please don't remove it
2024-06-20T16:28:02.921Z
<Patrick Donnelly> at least not yet
2024-06-20T16:40:08.395Z
<Adam Kraitman> 👍
2024-06-20T16:51:33.628Z
<Laura Flores> Thanks for bringing this up @Stefan Kooman. As we went over in the User + Dev meetup, adding your crush rule and osdmap to the tracker will be helpful in adding that test case.
2024-06-20T17:26:32.702Z
<Patrick Donnelly> this is so weird, the changes I made in 58174 were working and not fail with a puzzling redmine exception
2024-06-20T17:26:38.516Z
<Patrick Donnelly> adam did you change anything in the last hour?
2024-06-20T17:28:50.508Z
<Patrick Donnelly> @Adam Kraitman ^
2024-06-20T17:31:12.070Z
<Patrick Donnelly> nvm
2024-06-20T17:47:13.496Z
<Adam Kraitman> @Patrick Donnelly in the last hour I added access to the developers group on the tags
2024-06-20T17:51:57.851Z
<Patrick Donnelly> k
2024-06-20T17:52:03.396Z
<Patrick Donnelly> i'm not sure what happened but I think I fixed it

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