From: Patrick Donnelly <pdonnell@redhat.com>
Date: Wed, 12 Sep 2018 01:49:18 +0000 (-0700)
Subject: qa: tolerate failed rank while waiting for state
X-Git-Tag: v14.0.1~145^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=e52976e64d04d97910582f03185832e0ab2ea009;p=ceph.git

qa: tolerate failed rank while waiting for state

Fixes: http://tracker.ceph.com/issues/35828

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
---

diff --git a/qa/tasks/cephfs/filesystem.py b/qa/tasks/cephfs/filesystem.py
index b2837c2eee82..c0bbb20bb324 100644
--- a/qa/tasks/cephfs/filesystem.py
+++ b/qa/tasks/cephfs/filesystem.py
@@ -947,9 +947,17 @@ class Filesystem(MDSCluster):
         while True:
             status = self.status()
             if rank is not None:
-                mds_info = status.get_rank(self.id, rank)
-                current_state = mds_info['state'] if mds_info else None
-                log.info("Looked up MDS state for mds.{0}: {1}".format(rank, current_state))
+                try:
+                    mds_info = status.get_rank(self.id, rank)
+                    current_state = mds_info['state'] if mds_info else None
+                    log.info("Looked up MDS state for mds.{0}: {1}".format(rank, current_state))
+                except:
+                    mdsmap = self.get_mds_map(status=status)
+                    if rank in mdsmap['failed']:
+                        log.info("Waiting for rank {0} to come back.".format(rank))
+                        current_state = None
+                    else:
+                        raise
             elif mds_id is not None:
                 # mds_info is None if no daemon with this ID exists in the map
                 mds_info = status.get_mds(mds_id)