[LU-2787] review-zfs lustre-initialization_1 It is broken.

• Status: Resolved
• Project: Lustre
• Type: Bug
• Priority: Major
• Reporter: Keith Mannthey (Inactive)
• Assignee: WC Triage
• Resolution: Duplicate

Environment

a patch pushed via git.

• Severity: 3

Description

This is reported in every single patch review.

I didn't find an LU of this common error so I filed one. It seems very serious.

Perhaps this is a TT issue I don't know. I took a quick look at the system logs and all I saw were systems just sitting around for some time.

Actual error logs can be seen here:
https://maloo.whamcloud.com/test_sets/216df032-71f8-11e2-aad1-52540035b04c

The tests reports:
Failure Rate: 44.00% of last 100 executions [all branches]

Issue Links

• duplicates: LU-2469 Test-framework.sh ignores MDSDEV when...

Comments

[Andreas Dilger]

Keith,
you are right that this is a fairly serious problem, which should hopefully be fixed soon. I believe that there are a number of tickets tracking ZFS problems already, and Nathaniel and Li Wei are working to get ZFS running again. Unfortunately, since we weren't doing ZFS testing all the time, a number of bugs were introduced that now need to be fixed.

In order to make this a useful bug report, you need to look into the maloo test logs to see what the root of the problem is. Normally the "autotest" log is interesting for lustre-initialization-1 failures. In this case it shows near the end:

https://maloo.whamcloud.com/test_logs/2401a938-71f8-11e2-aad1-52540035b04c

Invalid filesystem name /dev/lvm-MDS/P1

which is LU-2469 and there is a patch at http://review.whamcloud.com/5016.

[Keith Mannthey (Inactive)]

Thanks for the pointers to the autotest log. I had looked at system, dmesg, and console logs. lustre-initialization-1 falls into TT issue for the most part for me, I will be more through in the future. In general I only looked at the ZFS logs because you have to dig into the maloo logs to even find out it was the ZFS run (there is an open TT for this issue.)

I was just wanting to be sure there was an LU placeholder for the issue. I searched for open LU's but I didn't seem to find LU-2469, it is great to see all the hard work stabilizing ZFS in autotest.