[LU-6840] update memory reply data in DNE update replay

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.8.0
• Affects Version/s: Lustre 2.8.0
• Severity: 3
• Assignee: Di Wang
• Reporter: Di Wang

Description

DNE update replay might update last_rcvd file, but because it will be operated on OSD/OSP directly, so the memory structure (ted/lrd etc) will not be updated by update replay handler, so it needs update these memory structure by itself after each update replay. (see replay_request_or_update() ).

In current implementation, this is done in target_update_lcd(), but multiple slot patch just changed this process, so target_update_lcd needs to be fixed as well.

Issue Links

• is blocking: LU-6773 DNE2 Failover and recovery soak testing
• is related to: LU-6831 The ticket for tracking all DNE2 bugs
• is related to: LU-6844 replay-single test 70b failure: 'rundbench load on * failed!'
• is related to: LU-7077 Pointer 'hash' returned from call to function 'cfs_hash_getref' may be NULL and will be dereferenced.
• is related to: LU-5319 Support multiple slots per client in last_rcvd file

Comments

[Joseph Gmitter (Inactive)]

Landed for 2.8.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/15576/
Subject: LU-6840 target: update reply data after update replay
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 9543df37cdfd35980c888440265d161e350d166d

[Di Wang]

For cross-MDT operation, the updates (including update last_rcvd and reply_data) will be recorded on all of MDTs. So during the recovery, once the updates on the master MDT are missing, then these updates needs to be redo on the master MDT, which means these last_rcvd and reply_data file needs to be updated as well in this process, and those memory stuff (lcd etc) needs to be updated as well, which is normally through the trans stop callback.

In current implementation, due to the speciality of the DNE recovery and recent changes of multiple slot, stop callback can not be called correctly, which is this patch trying to resolve.

[Gregoire Pichon]

I am not sure to understand the context, but the metadata operations operated on OSD/OSP should not make use of the multiple slot patch.

At the moment (lustre 2.8) multiple slot feature is only supported for MDT exports that have the OBD_CONNECT_MULTIMODRPCS flag, that is to say exports for MDC. This flag should not be set on exports for OSP.

[Gerrit Updater]

wangdi ([e-mail]) uploaded a new patch: http://review.whamcloud.com/15576
Subject: LU-6840 target: update reply data after update replay
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 49d65b184daf7c6d580f918b3befd41a16c1f278