Bug #50140: test/thrash - scrub: "not all pgs scrubbed" due to short rescrubbing period - RADOS - Ceph

Actions

Copy link

Bug #50140

closed

test/thrash - scrub: "not all pgs scrubbed" due to short rescrubbing period

Added by Ronen Friedman about 3 years ago. Updated about 3 years ago.

Status:

Duplicate

Priority:

Normal

Assignee:

David Zafman

Category:

Tests

Target version:

% Done:

Source:

Tags:

Backport:

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

rados

Component(RADOS):

OSD

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

Thrashing test:
The observed error is "Exiting scrub checking -- not all pgs scrubbed".

See below for analysis.

Related issues 1 (1 open — 0 closed)

Actions

Copy link

Updated by Ronen Friedman about 3 years ago

Assignee set to David Zafman

Caused by a combination of:
- re-scrub period ("osd scrub min interval") is set in radod/thrash* to (only) 60s.
- a large set of PGs to scrub.
- a PG that failed to reserve replica resources.

The failure flag will only be erased once the queue of PGs to scrub is empty. But under the
first two conditions - that never happens.

Actions

Copy link

Updated by Ronen Friedman about 3 years ago

Possible fixes to consider:

- a simple fix: extending the tests min-scrub-time;
- possibly better: modify the handling of the "failed once in achieving replicas' resources"
to be periodically cleared.

Actions

Copy link