[LU-10948] client cache open lock after N opens

• Type: Improvement
• Resolution: Unresolved
• Priority: Minor
• Affects Version/s: Lustre 2.9.0
• Environment: cent server/sles client

Epic/Theme

Performance

• Assignee: Oleg Drokin
• Reporter: Bob Ciotti (Inactive)

Description

listed as minor but when a user does this, we start to get phone calls form other users and then page POC to identify code/user. Workaround is to terminate user job(s).

Oleg has said that Lustre has an existing feature for a client to acquire open lock but off by default. This to mimic NFS behavior.

Ideal change would be that we can specify a number of times that a file is opened on single client at which time lock is acquired. (e.g. 10th time)

Use case is naive user who loop in this way on like 5000+ threads in java:

do until till the sun turns black()

{ fd = open(*my_thread_ID, O_APPEND) calculate_something_small_but_useful() write(fd, *fortytwo, 42) close(fd }

Users often don't have complete control over the code they run and as a result may not be able to quickly make even simple changes.

Issue Links

• is duplicated by: LU-5426 Add more controls for open file cache
• is related to: LU-12262 Improve sbi_flags checking
• is related to: LU-14694 racer timeouts with LU-10948
• is related to: LU-14743 Interop: sanity test 429 fails with 'multiop failed'
• is related to: LU-16463 llapi_open_by_fid open should not be cached systematically
• is related to: LU-10957 Return more lock bits from MDS for open-create
• is related to: LU-11623 Allow caching of open-created dentries
• is related to: LU-7915 investigate heuristics for SPARK client getting MDS openlock
• is related to: LU-4398 mdt_object_open_lock() may not flush conflicting handles
• is related to: LU-10269 Fixes for selective trybits
• is related to: LU-11199 mdsrate open() performance degradation
• is related to: LU-8585 All Lustre test suites should pass with subdirectory mount
• is related to: LU-10955 Cancel open lock as part of close request
• is related to: LU-7915 investigate heuristics for SPARK client getting MDS openlock

Comments

[Andreas Dilger]

Patch https://review.whamcloud.com/32156 "LU-10948 mdt: Always return lookup lock on opens" is not landed yet.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/43907/
Subject: LU-10948 mdt: New connect flag for non-open-by-fid lock request
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 72c9a6e5fb6e11fca1b1438ac18f58ff7849ed7d

[Gerrit Updater]

Oleg Drokin ([e-mail]) uploaded a new patch: https://review.whamcloud.com/43907
Subject: LU-10948 mdt: New connect flag for non-open-by-fid lock request
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f6acd02aa06420264b83e4ef2d4bf2f0f95794c7

[Gerrit Updater]

Oleg Drokin ([e-mail]) uploaded a new patch: https://review.whamcloud.com/43903
Subject: LU-10948 mdt: New connect flag for non-open-by-fid lock request
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 14edc80c617871e550d41a0f2e0b0f1491e02f37

[Jay Lan (Inactive)]

Can you port #32158 to b2_12? Thanks!

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/32158/
Subject: LU-10948 llite: Introduce inode open heat counter
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 41d99c4902836b7265db946dfa49cf99381f0db4

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/41091/
Subject: LU-10948 llite: Revalidate dentries in ll_intent_file_open
Project: fs/lustre-release
Branch: b2_12
Current Patch Set:
Commit: 6ad2679352ee9a14eceb8c70b37df084bc07a6f7

[Gerrit Updater]

Mike Pershin ([e-mail]) uploaded a new patch: https://review.whamcloud.com/41091
Subject: LU-10948 llite: Revalidate dentries in ll_intent_file_open
Project: fs/lustre-release
Branch: b2_12
Current Patch Set: 1
Commit: d1242a54f369712b5c39cc51e3118e9f83c5895e

[Andreas Dilger]

This issue has not been resolved, as the main functional patches have not been landed yet. The few patches that have been landed so far are just cleaning up issues in the code.

There are two patches from this ticket outstanding for this ticket that add a simple client-side per-inode counter+threshold to decide whether the client should fetch the openlock:

• https://review.whamcloud.com/32157 "LU-10948 llite: Revalidate dentries in ll_intent_file_open"
• https://review.whamcloud.com/32158 "LU-10948 llite: WIP! Introduce inode open counter"

There is also an older patch on LU-7915 that just provides a whole-client tunable to enable opencache:

• https://review.whamcloud.com/19664 "LU-7915 llite: add a tunable to enable opencache"

Getting the patches for this ticket updated to the latest master and reviewing/investigating the test failures would be useful for getting it landed sooner.

The LU-7915 patch might be useful for testing how much opencache would help a specific workload, but always-on opencache had shown to be a performance hit in some cases in the past, which is why we didn't enable opencache permanently. Still, landing patch 19664 would at least give an available tunable to turn this on/off on a per-client basis if it shows a net improvement for a site's workload.

[Alex Kulyavtsev]

Is this resolved?

In what lustre release this patch is available?