Open xhernandez opened 1 year ago
Any updates on this one ? newer runs still don't have anything inside var/log/glusterfs and I can't analyze regression failures.
Seems to still be a issue. That's curious, because nothing changed recently.
Ok so checking the last build, it seems that something remove the log, then create the tarball.
See at the end of https://build.gluster.org/job/gh_centos7-regression/3200/console
16:33:50 + rm -rf /var/lib/glusterd/groups /var/lib/glusterd/hooks /var/log/glusterfs/.cmd_log_history '/etc/glusterd/*' '/var/log/glusterfs/*' /build/export
then
16:33:52 sudo -E tar -czf "${filename}" /var/log/glusterfs /var/log/messages*;
That's here since 2014:
So I am not sure why it was noticed only today. But the fix is to change that to not clean the log.
That's a bit weird because I'm pretty sure I've downloaded many archives during these years and the contents were there.
Just a recent example: https://build.gluster.org/job/gh_centos7-regression/3195/
This one contains data, though not everything it should be there.
In any case, the code should be changed.
Ok so sometime, it work, and sometime it doesn't. Exactly the kind of problem I love to deal with ( sarcasm ).
I remember we had 2 different types of failure on jenkins, and I guess depending on the type of issue, it could go a different code path.
I'm not sure if it could help but other jobs, like centos8-regression
, seem to always generate the tgz correctly (though the test itself fails constantly).
The build artifact glusterfs-logs.tgz that is generated during the execution of jenkins' job gh_centos7-regression only contains /var/log/messages. Inside /var/log/glusterfs there's only a single file named '.notar'.
Some examples: