qa/cephfs: add probabilistic ignorelist for pg_health

PG_AVAILABILITY/PG_DEGRADED warnings are dominating fs runs. We want the
underlying issue fixed but it cannot continue to fail all of our tests 100% of
the time. Use a probabilistic addition of these warnings to the ignorelist.

Fixes: https://tracker.ceph.com/issues/64984
Related-to: https://tracker.ceph.com/issues/52624
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

diff --git a/qa/cephfs/overrides/pg_health.yaml b/qa/cephfs/overrides/pg_health.yaml
new file mode 100644 (file)
index 0000000..defd70e
--- /dev/null
+++ b/qa/cephfs/overrides/pg_health.yaml
@@ -0,0 +1,14 @@
+# To prevent PG_AVAILABILITY failures from dominating runs, we usually ignore
+# these warnings. However, we'd like to see the cause of these warnings to be
+# fixed so occasionally allow the warning to fail the job.
+#
+# https://tracker.ceph.com/issues/52624
+
+teuthology:
+  premerge: if math.random() < 0.05 then reject() end
+overrides:
+  ceph:
+    log-ignorelist:
+      - PG_AVAILABILITY
+      - PG_DEGRADED
+      - Reduced data availability