[LU-2277] Text file busy

• Type: Bug
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.1.5
• Affects Version/s: Lustre 2.1.3
• Environment: server-2.1.3-1nas (centos6.3); client 2.1.3-1nas Suse11-1
• Severity: 3
• Assignee: Nathaniel Clark
• Reporter: Mahmoud Hanafi

Description

1. ldd a.out
   /usr/bin/ldd: line 118: ./a.out: Text file busy

see attached client and server debug traces.

Issue Links

• duplicates: LU-1571 It should not update last xid for replay open.
• duplicates: LU-971 Re-sent open requests cause mod_count leakages

Comments

[Nathaniel Clark]

Patches picked to branch

[Nathaniel Clark]

Original Commit is now just testing updates:

Following are cherry-picked commits that are dependent on 4848, 5002 is unchanged, 5003 needed a merge conflict resolved.
http://review.whamcloud.com/5002
http://review.whamcloud.com/5003

[Nathaniel Clark]

http://review.whamcloud.com/4848

[Nathaniel Clark]

Client seems to be re-sending open create because response data is large and reply_in_callback() registers it as truncated and resends. So the solution I'm going to look to implement is to correctly handle the replay of the open/create for write/exec in mdt_mfd_open().

[Nathaniel Clark]

Note on what goes over the wire:
Items listed are wireshark dump of traffic, all items are requests client -> mds, all items are replied to prior to the next item being sent.

For stripe sizes 50-53: (example is stripe 50)
1) LDLM_ENQUEUE request [Concurrent Write][ intent : open create ] filename : a2
2) MDS_CLOSE [REINT_SETATTR]
3) LDLM_ENQUEUE request [Concurrent Read][ intent : lookup ] filename : 50
4) LDLM_ENQUEUE request [Concurrent Read][ intent : open ][REINT_OPEN] filename : a2
5) MDS_CLOSE [REINT_SETATTR]

For stripe sizes 54-59: (example is stripe 54)
1) LDLM_ENQUEUE request [Concurrent Write][ intent : open create ] filename : a2
2) LDLM_ENQUEUE request [Concurrent Write][ intent : open create ] filename : a2
3) MDS_CLOSE [REINT_SETATTR]
4) LDLM_ENQUEUE request [Concurrent Read][ intent : lookup ] filename : 54
5) LDLM_ENQUEUE request [Concurrent Read][ intent : open ] filename : a2
6) LDLM_ENQUEUE request [Concurrent Read][ intent : open ] filename : a2
7) MDS_CLOSE [REINT_SETATTR]