search

osbuild / osbuild-composer

An HTTP service for building bootable OS images.
https://www.osbuild.org
Apache License 2.0
156 stars 107 forks source link

Fix executor setting its own hostname #4229

Closed croissanne closed 3 weeks ago

croissanne commented 3 weeks ago

Cloud-init data was broken + the script wasn't executable :facepalm:

ezr-ondrej commented 3 weeks ago

Failing on Fedora 40 if I'm reading it correctly.

jrusz commented 3 weeks ago

There is something wrong with the filesystem.sh test on fedora-40, it seem image-info has issue there, I'd wait with the transition for now to fix the stage runners.

ezr-ondrej commented 3 weeks ago

Yeah I'd agree, otherwise we'd need to get it in broken to fix stage, right?

jrusz commented 3 weeks ago

Yeah I'd agree, otherwise we'd need to get it in broken to fix stage, right?

Well breaking main is not desirable so I'd say otherwise we'd need to debug and fix the issue which would delay the fix for stage :)

croissanne commented 3 weeks ago

There is something wrong with the filesystem.sh test on fedora-40, it seem image-info has issue there, I'd wait with the transition for now to fix the stage runners.

Can't run fedora-38 anymore since the rpmrepo snapshots got deleted, so main currently is broken already.

jrusz commented 3 weeks ago

There is something wrong with the filesystem.sh test on fedora-40, it seem image-info has issue there, I'd wait with the transition for now to fix the stage runners.

Can't run fedora-38 anymore since the rpmrepo snapshots got deleted, so main currently is broken already.

Yeah well in that case I'd just not run it for now and fix in a followup... I read the comments and it seems there was something missing in the kernel, we should check if it's there already.

croissanne commented 3 weeks ago

There is something wrong with the filesystem.sh test on fedora-40, it seem image-info has issue there, I'd wait with the transition for now to fix the stage runners.

Can't run fedora-38 anymore since the rpmrepo snapshots got deleted, so main currently is broken already.

Yeah well in that case I'd just not run it for now and fix in a followup... I read the comments and it seems there was something missing in the kernel, we should check if it's there already.

the issue now seems to be image-info:

Traceback (most recent call last):
  File "/usr/libexec/osbuild-composer-test/image-info", line 2891, in <module>
    main()
  File "/usr/libexec/osbuild-composer-test/image-info", line 2885, in main
    report = analyse_image(target)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/libexec/osbuild-composer-test/image-info", line 2731, in analyse_image
    append_partitions(report, target)
  File "/usr/libexec/osbuild-composer-test/image-info", line 2706, in append_partitions
    append_filesystem(report, root_tree)
  File "/usr/libexec/osbuild-composer-test/image-info", line 2317, in append_filesystem
    dnf_conf = read_dnf_conf(tree)
               ^^^^^^^^^^^^^^^^^^^
  File "/usr/libexec/osbuild-composer-test/image-info", line 2158, in read_dnf_conf
    dnf_config = _read_inifile_to_dict(f"{tree}/etc/dnf/dnf.conf")
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/libexec/osbuild-composer-test/image-info", line 2124, in _read_inifile_to_dict
    parser.readfp(f)
    ^^^^^^^^^^^^^
AttributeError: 'RawConfigParser' object has no attribute 'readfp'. Did you mean: 'read'?
ERROR image-info failed, show last few kernel message to debug
ezr-ondrej commented 3 weeks ago

Are we ok with being out of date, can someone force rebase merge if the tests pass?

croissanne commented 3 weeks ago

Are we ok with being out of date, can someone force rebase merge if the tests pass?

composer doesn't require prs to be rebased becuse of CI taking a long time. But yes I can always force merge too

croissanne commented 3 weeks ago

filesystem.sh succeeded + WorkerExecutor did, merging to get the worker image build rolling, will follow up with an image-info fix.