[LU-3787] replay-single test_90: wrong stripe: f2, uuid: lustre-OST0002_UUID Created: 20/Aug/13  Updated: 03/Sep/13  Resolved: 03/Sep/13

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.5.0
Fix Version/s: Lustre 2.5.0

Type: Bug Priority: Minor
Reporter: James Nunez (Inactive) Assignee: James Nunez (Inactive)
Resolution: Fixed Votes: 0
Labels: revzfs

Issue Links:
Related
is related to LU-3809 replay-single test_58c does not clean... Resolved
Rank (Obsolete): 9794

 Description   

This issue relates to the following test suite run: http://maloo.whamcloud.com/test_sets/b034018c-0756-11e3-bc50-52540035b04c.

The sub-test test_90 failed with the following error:

wrong stripe: f2, uuid: lustre-OST0002_UUID

From the client test log, we see

== replay-single test 90: lfs find identifies the missing striped file segments == 02:30:16 (1376731816)
Create the files
error opening /mnt/lustre/d90.replay-single/file: No such file or directory (2)
llapi_semantic_traverse: Failed to open '/mnt/lustre/d90.replay-single/file': No such file or directory (2)
error: getstripe failed for /mnt/lustre/d90.replay-single/file.
 replay-single test_90: @@@@@@ FAIL: wrong stripe: f2, uuid: lustre-OST0002_UUID 

So, the real error is that there is no file by the name of /mnt/lustre/d90.replay-single/file.

Look at the code for test 90, there is a missing "$" in front of "file" when calling getstripe. I'll upload a patch.



 Comments   
Comment by James Nunez (Inactive) [ 20/Aug/13 ]

Proposed patch is at:

http://review.whamcloud.com/7403

Comment by Oleg Drokin [ 21/Aug/13 ]

So the patch, while good, does not address the actual issue of the file not having the striping requested (and the bug you fix is then in the error path).

Now, I wonder if what happened here is MDS is out of precreated objects for this OST and so it picks some other OST (we used to strictly enforce that requesting striping matches the actual one, but that was related some time ago, right Alex?)
If this is the case (needs to be verified in the code) then the fix might be to make the test to ensure MDS has big enough pool of preallocated tests before starting with the file creation.

Comment by James Nunez (Inactive) [ 03/Sep/13 ]

Landed to master.

Generated at Sat Feb 10 01:36:54 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.