Bug #61950
openmds/OpenFileTable: match MAX_ITEMS_PER_OBJ does not honor osd_deep_scrub_large_omap_object_key_threshold
0%
Description
The changes implemented in [1] should make sure that we never have an openfiletable objects omap keys above osd_deep_scrub_large_omap_object_key_threshold( default 200000) but in Ceph version: 14.2.22-110.el8cp (2e0d97dbe192cca7419bbf3f8ee6b7abb42965c4) nautilus
We see the following message in `ceph -s` and the following log in cluster logs:
LARGE_OMAP_OBJECTS
2023-07-09 10:42:34.177821 osd.123 (osd.123) 16631 : cluster [WRN] Large omap object found. Object: 3:2adbd678:::mds0_openfiles.9:head PG: 3.1e6bdb54 (3.154) Key count: 200001 Size (bytes): 12824872
In this situation the mds0_openfiles.9 object has just 1 key extra from the threshold(200000).
[1] Old Tracker and PR:
https://tracker.ceph.com/issues/42515
https://github.com/ceph/ceph/pull/31232
Updated by Venky Shankar 10 months ago
Ok, so an off-by-one error - should be relatively easy to figure...
Updated by Patrick Donnelly 10 months ago
- Category set to Correctness/Safety
- Status changed from New to In Progress
- Assignee set to Patrick Donnelly
- Priority changed from Normal to High
- Target version set to v19.0.0
- Source set to Support
- Backport set to reef,quincy,pacific
- Component(FS) MDS added
Updated by Venky Shankar 10 months ago
Patrick - I can take this one if you haven't started on it yet.
Updated by Patrick Donnelly 10 months ago
Venky Shankar wrote:
Patrick - I can take this one if you haven't started on it yet.
I have started on it. Should be quick.
Updated by Patrick Donnelly 10 months ago
- Related to Bug #42515: fs: OpenFileTable object shards have too many k/v pairs added
Updated by Patrick Donnelly 10 months ago
Venky Shankar wrote:
Patrick - I can take this one if you haven't started on it yet.
https://github.com/ceph/ceph/pull/52415#issuecomment-1632803909
I haven't found the cause. If you'd like to have a stab at it be my guest. :)
Updated by Venky Shankar 8 months ago
Patrick Donnelly wrote:
Venky Shankar wrote:
Patrick - I can take this one if you haven't started on it yet.
https://github.com/ceph/ceph/pull/52415#issuecomment-1632803909
I haven't found the cause. If you'd like to have a stab at it be my guest. :)
I'll have a look. Assigning to myself.
Updated by Venky Shankar 8 months ago
- Assignee changed from Patrick Donnelly to Venky Shankar
Updated by Patrick Donnelly 5 months ago
- Status changed from In Progress to Need More Info
- Assignee deleted (
Venky Shankar) - Pull request ID deleted (
52415)
cannot reproduce