[LU-2990] Failure on sanity test_24v: error in listing large dir

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.4.0
• Affects Version/s: Lustre 2.4.0
• Labels: MB zfs
• Environment: server and client: lustre-master build# 1328 RHEL6; fstype is zfs
• Severity: 3
• Assignee: nasf (Inactive)
• Reporter: Sarah Liu

Description

https://maloo.whamcloud.com/test_sessions/a84605f2-90c2-11e2-8311-52540035b04c

client console:

Lustre: DEBUG MARKER: == sanity test 24v: list directory with large files (handle hash collision, bug: 17560) == 10:15:47 (1363713347)
Lustre: DEBUG MARKER: cancel_lru_locks mdc start
Lustre: DEBUG MARKER: cancel_lru_locks mdc stop
Lustre: DEBUG MARKER: sanity test_24v: @@@@@@ FAIL: error in listing large dir
Lustre: 8261:0:(dir.c:463:ll_get_dir_page()) Page-wide hash collision: 5723140463788032
LustreError: 8261:0:(dir.c:594:ll_dir_read()) error reading dir [0x200000400:0x4:0x0] at 5723140463788032: rc -5
Lustre: 8261:0:(dir.c:463:ll_get_dir_page()) Page-wide hash collision: 5723140463788032
LustreError: 8261:0:(dir.c:594:ll_dir_read()) error reading dir [0x200000400:0x4:0x0] at 5723140463788032: rc -5

Issue Links

• is duplicated by: LU-3119 System hang when running sanity test 24x on DNE with ZFS
• is related to: LU-3605 Sanity test suite aborts in test 24v

Comments

[Peter Jones]

Landed for 2.4

[nasf (Inactive)]

This is the patch:

http://review.whamcloud.com/#change,5894

[Andreas Dilger]

We shouldn't get a hash collision with just 100k files. I wonder if something bad is happening with the hash mapping in the ZFS code?

[Jodi Levi (Inactive)]

Fan Yong,
Could you please have a look at this one?
Thank you!

[Sarah Liu]

Here are the debug log and trace from the client. The system actually hung, so I abort the testing.

[Keith Mannthey (Inactive)]

The error logs for sanity show ABORT and do not contain any logs for test24v. https://maloo.whamcloud.com/test_sets/aab26060-90c2-11e2-8311-52540035b04c

I do not seem to be able to search in maloo for sanity test_24v. It seems there is a test_24u and test_24w but no v.

Do you have more info you can share? How did you run the test?