[LU-873] IOR single shared file test fails

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.2.0, Lustre 1.8.7
• Type: Bug
• Priority: Major
• Reporter: Cliff White (Inactive)
• Assignee: Zhenyu Xu
• Resolution: Cannot Reproduce

Environment

Hyperion/LLNL

• Severity: 3

Description

IOR fails when > 10 clients are run.

0289: ERROR in aiori-POSIX.c (line 256): transfer failed.
0289: ERROR: Input/output error
0289: ** exiting **
0289: [289] [MPI Abort by user] Aborting Program!
0289: [289:hyperion355] Abort: MPI_Abort() code: -1, rank 289, MPI Abort by user Aborting program ! at line 99 in file mpid_init.c
https://maloo.whamcloud.com/test_sessions/53fb3ae2-155a-11e1-b669-52540025f9af

Comments

[Peter Jones]

Bobi

Could you please look into this one?

Thanks

Peter

[Cliff White (Inactive)]

This may be related/identical to LU-873, filed by LLNL.

[Jinshan Xiong (Inactive)]

I took a look at the log. It looks like the client hyperion360 was requesting an RW lock with local cookie: 0x389db5cc182f83cd from OST0000. However, the completion RPC was dropped(client never got this RPC) so that the status of this lock on the server is granted, but client kept waiting completion; then this lock on the client will never be revoked because a process is waitinig for it. This is just my guess because the log was truncated.

Please also notice that completion AST is not resendable; also there are several bulk transfer error on the console. I don't know if network ran into problem at that time.

[Christopher Morrone]

LLNL's ticket is LU-874.

[Andreas Dilger]

Close old ticket.