action #36099
closed
[tools] SKIPTO doesn't work anymore: Failed creating CGroup subtree
Added by dheidler almost 6 years ago.
Updated almost 6 years ago.
Category:
Regressions/Crashes
Description
I ran a test with
MAKETESTSNAPSHOTS=1
and tried to skip to a certain module using
SKIPTO=x11-multi_users_dm
in a 2nd run.
This error appeared in worker.log and the test failed:
Use of uninitialized value in subroutine entry at /usr/lib/perl5/vendor_perl/5.26.1/Mojo/File.pm line 84.
[error] Failed creating CGroup subtree 'mkdir /sys/fs/cgroup/0: Read-only file system at /usr/lib/perl5/vendor_perl/5.26.1/Mojo/File.pm line 72.
', disabling them. You can define a custom slice with OPENQA_CGROUP_SLICE or indicating the base mount with MOJO_CGROUP_FS
I run the openqa worker as my local user.
- Subject changed from SKIPTO doesn't work anymore: Failed creating CGroup subtree to [tools] SKIPTO doesn't work anymore: Failed creating CGroup subtree
- Category set to Regressions/Crashes
- Description updated (diff)
- Priority changed from High to Normal
So what if you don't run it as your user but within the systemd service it's supposed to use?
Apart from the face that loading a snapshot from a qcow2 fiel shouldn't require some cgroup foo, the systemd unit file also doesn't run the worker as root.
No, but within a cgroup slice. So just answer my question - because if it's broken in your personal setup or in general is a crucial information.
I don't use the systemd unit file anymore so that I can run it easily from git.
Can you link to the test?
If you are starting the worker normally, as an user from bash (no systemd involved), it should still work. (Otherwise, it wouldn't work here: http://phobos.suse.de/tests/11092#settings as I also run this from git)
qemu-img snapshot -l /srv/data/openqa/development_basedir/openqa/pool/1/raid/1
Snapshot list:
ID TAG VM SIZE DATE VM CLOCK
1 install-boot 5.6M 2018-04-17 17:18:20 00:00:00.087
2 install-fix_repos 1.5G 2018-04-17 17:19:00 00:00:39.967
3 install-openqa_webui 1.8G 2018-04-17 17:19:51 00:01:29.964
8 osautoinst-worker 1.9G 2018-04-18 13:56:16 00:07:23.575
10 install-test_distribution 1.9G 2018-04-18 18:43:28 00:04:33.724
11 openQA-dashboard 1.9G 2018-04-18 18:43:29 00:04:34.066
12 openQA-login 1.9G 2018-04-18 18:43:49 00:04:52.317
13 openQA-build_results 1.9G 2018-04-18 18:43:52 00:04:54.936
14 install-openqa_worker 1.9G 2018-04-18 19:02:42 00:03:48.892
18 parser-all 1.9G 2018-04-26 12:31:27 00:04:41.425
19 shutdown-shutdown 1.9G 2018-05-07 20:41:08 00:06:13.673
In any case, the message you are seeing is just isotovideo telling you that it's disabling CGroups as it cannot create any (For subreapers/sessions/stuff)
- Status changed from New to Closed
[dheidler@artemis 2]$ pwd
/home/dheidler/.local/lib/openqa/pool/2
[dheidler@artemis 2]$ qemu-img snapshot -l raid/l1
Snapshot list:
ID TAG VM SIZE DATE VM CLOCK
1 boot-boot_to_desktop 3.3M 2018-05-11 14:29:53 00:00:00.011
3 console-consoletest_setup 927M 2018-05-11 14:30:57 00:01:03.058
5 console-hostname 950M 2018-05-11 14:32:46 00:02:49.748
6 lastgood 950M 2018-05-11 14:33:17 00:03:18.830
7 x11-chrome 950M 2018-05-11 14:33:19 00:03:18.846
8 x11-multi_users_dm 956M 2018-05-11 14:34:58 00:04:56.812
9 x11-wine 951M 2018-05-11 14:38:20 00:08:17.429
I was just able to start a job with skipto - I suspect that due to my multimachine setup with two local workers the job got scheduled on the wrong worker wich didn't have the snapshots in its qcow2 file.
http://artemis.suse.de/tests/289
Maybe the cgroup error message should be converted into a warning as it is not fatal and - I suppose - only for performance.
Also available in: Atom
PDF