2025-01-07T18:37:43.010Z

<John Mulligan> Hi team, we are thinking about enhancing the way cephadm can be used to debug ceph. In particular, we heard feedback at cephalocon around running gdb and tools like that. We're curious if anyone has tried the
debugging documentation at <https://docs.ceph.com/en/squid/cephadm/troubleshooting/#running-the-debugger-with-cephadm>  and if so how much value there would be in automating some of these actions in the cephadm command. Any other thoughts on this topic are welcome too.

(Please reply in thread if possible, thanks!)

2025-01-07T19:13:49.093Z

<yuriw> 4 - run-tox-mgr (Failed)

2025-01-07T19:14:09.469Z

<yuriw> I have several PRs failing like this

2025-01-07T19:14:23.178Z

<yuriw> <https://github.com/ceph/ceph/pull/61094>

2025-01-07T19:46:59.598Z

<Zac Dover> I saw some of these last week, so it's not an isolated phenomenon.

2025-01-07T20:41:39.587Z

<Christopher Hoffman> Seeing it here too: <https://github.com/ceph/ceph/pull/56894>

I tracked it down to a ceph-build change from Dec. See this comment: <https://github.com/ceph/ceph/pull/56894#issuecomment-2575490955>

2025-01-07T20:59:21.163Z

<John Mulligan> I was pinged by Naman last week about this. The failure I see in Yuri's PR is the same as the one Naman showed me.

2025-01-07T20:59:44.658Z

<John Mulligan> @Christopher Hoffman I don't see the connection between the tox (mypy) failure and the 2nd link you shared.

2025-01-07T21:00:03.118Z

<John Mulligan> This seems to be impacting reef (only?) backports

2025-01-07T21:00:49.274Z

<John Mulligan> I told Naman that I couldn't see any recent changes to the python code that could cause this.... so I figured the change was due to environment changing (python version probably, this failure doesn't appear on python 3.8)

2025-01-07T21:03:03.990Z

<John Mulligan> FWIW I could **not** reproduce the exact failure on my system, even using ubuntu container images. But it did go away when I ran it with ubuntu 20.04 (python 3.8)

2025-01-07T21:05:00.660Z

<Christopher Hoffman> Hmm, I saw these lines in Yuri's linked PR and similar output in the one I linked
```changed files:
qa/suites/rbd/migration/6-prepare/qcow2-https.yaml
src/librbd/migration/HttpClient.cc
++ echo 'qa/suites/rbd/migration/6-prepare/qcow2-https.yaml
src/librbd/migration/HttpClient.cc'
++ egrep -v '^container/'
++ wc -l
+ '[' 2 -gt 0 ']'
+ return 1```
The things that stood out to me was "changed files:" and then the lines with +s. Then I was able to track it down to potentially a change in ceph-build

2025-01-07T21:05:53.521Z

<John Mulligan> Interesting. I was focused on the `mypy` failure that causes the tox job to fail.

2025-01-07T21:06:18.607Z

<John Mulligan> I wonder if somehow it was already failing the mypy test but it wasn't catching it? But now it does?

2025-01-07T21:07:46.673Z

<John Mulligan> Dan's change is trying to avoid running the test suites if the changes only apply to the new container/ build dir.
I can point it out to him tomorrow if he doesn't see this thread on his own.

2025-01-07T21:08:12.467Z

<John Mulligan> Dan's change is trying to avoid running the test suites if the changes only apply to the new container/ build dir.  But it's not entirely clear to me how it works.
I can point it out to him tomorrow if he doesn't see this thread on his own.

2025-01-07T22:48:32.622Z

<Dan Mick> The change to avoid running tests if container/ only is just like the preexisting test for doc/; I can't imagine how this would be affecting mypy at all, much less in a non-obvious way

2025-01-07T22:51:09.971Z

<Dan Mick> is there something in the log showing how or why run-tox-mgr failed?  I can't see it