qa: add a YAML to ignore MGR_DOWN warning

RCA showed that it is not the NFS code that lead to the warning since the
warning occurred before the test cases started to execute, later on after
some discussion with the venky and greg, it was found that there were some
clog changes made recently which leads to this warning being added to the
clog.

Digging more further, it was found that the warning is generated when mgr fail
is run when there is no mgr available. The reason for unavailability is when
`setup_mgrs()` in class `MgrTestCase` stops the mgr daemons, sometimes the mgr
just crashes - `mgr handle_mgr_signal  *** Got signal Terminated ***`  and
after which `mgr fail` (again part of `setup_mgrs()`) is run and the `MGR_DOWN`
warning is generated.

This warning is only evident in nfs is because this is the only fs suite that
makes use of class `MgrTestCase`. To support my analysis, I had ran about eight
jobs in teuthology and I could not reproduce this warning. Since this is not
harming the NFS test cases execution and the logs do mention that the mgr
daemon did get restarted (`INFO:tasks.cephadm.mgr.x:Restarting mgr.x
(starting--it wasn't running)...`), it is good to conclude that ignoring this
warning is the simplest solution.

Fixes: https://tracker.ceph.com/issues/65265
Signed-off-by: Dhairya Parmar <dparmar@redhat.com>
(cherry picked from commit 7d954cefb1139011e90cd33e999aa357cbcdabb1)

diff --git a/qa/suites/fs/nfs/overrides/ignore_mgr_down.yaml b/qa/suites/fs/nfs/overrides/ignore_mgr_down.yaml
new file mode 100644 (file)
index 0000000..fb40742
--- /dev/null
+++ b/qa/suites/fs/nfs/overrides/ignore_mgr_down.yaml
@@ -0,0 +1,9 @@
+# When the NFS test class is constructed, the `MgrTestCase.setup_mgrs` invokes
+# `mgr fail` to restart the MGR which sometimes crashes the daemon and the
+# warning `MGR_DOWN` is generated. This is an intermittent failure which is
+# irrelevant to the NFS suite, and therefore should be ignored.
+
+overrides:
+  ceph:
+    log-ignorelist:
+      - MGR_DOWN