[LU-3662] Allow reading from ldlm.dump_namespaces or new ldlm.namespaces.*.dump_locks file Created: 30/Jul/13  Updated: 17/Jun/20

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.5.0
Fix Version/s: None

Type: Bug Priority: Minor
Reporter: John Hammond Assignee: WC Triage
Resolution: Unresolved Votes: 0
Labels: ldlm, medium

Issue Links:
Duplicate
Severity: 3
Rank (Obsolete): 9447

 Description   

Imperative lock dumping via writes to /proc/fs/lustre/ldlm/dump_namespaces only works once every 10 seconds because of the rate limitation imposed by ns_next_dump. Since ldlm_expired_completion_wait() has it's own 5 minute rate limit, it seems best just to do away with ns_next_dump altogether.

It would also be nice if we had a per namespace dump trigger and if we could just cat that file to get the dump rather than having to get it from the logs.



 Comments   
Comment by Andreas Dilger [ 14/May/20 ]

Being able to read the "dump_namespace" file directly with "lctl get_param" (via debugfs, similar to lite.*.dump_pages) instead of dumping it to the debug log and having to read it from there would be way better.

Generated at Sat Feb 10 01:35:50 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.