From: Sridhar Seshasayee Date: Fri, 9 Feb 2024 09:11:19 +0000 (+0530) Subject: qa: Add benign cluster warning from ec-inconsistent-hinfo test to ignorelist X-Git-Tag: v18.2.4~122^2 X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=0679a86300c247fc2a64f0794e4de631b16b3b13;p=ceph.git qa: Add benign cluster warning from ec-inconsistent-hinfo test to ignorelist The changes introduced in PR: https://github.com/ceph/ceph/pull/53524 made the randomized values of osd_op_queue and osd_op_queue_cut_off consistent across all OSD shards. Due to the above, ec-inconsistent-hinfo test could fail with the following cluster warning (benign) depending on the randomly selected scheduler type. "cluster [WRN] Error(s) ignored for 2:ad551702:::test:head enough copies available" In summary, the warning is generated due to the difference in the PG deletion rates between WPQ and mClock schedulers. Therefore, the warning shows up in cases where the mClock scheduler is the op queue scheduler chosen randomly for the test. The PG deletion rate with mClock scheduler is quicker compared to the WPQ scheduler since it doesn't use sleeps between each delete transaction and relies on the cost of the deletion which in turn is proportional to the average size of the objects in the PG. For a more detailed analysis, see the associated tracker. Fixes: https://tracker.ceph.com/issues/64573 Signed-off-by: Sridhar Seshasayee (cherry picked from commit 9677ad6dddf34302e5ad1cde6e72342d06e73642) --- diff --git a/qa/suites/rados/singleton/all/ec-inconsistent-hinfo.yaml b/qa/suites/rados/singleton/all/ec-inconsistent-hinfo.yaml index 31724f9e81b11..84abb702cf460 100644 --- a/qa/suites/rados/singleton/all/ec-inconsistent-hinfo.yaml +++ b/qa/suites/rados/singleton/all/ec-inconsistent-hinfo.yaml @@ -30,6 +30,7 @@ tasks: - slow request - unfound - \(POOL_APP_NOT_ENABLED\) + - enough copies available conf: osd: osd min pg log entries: 5