mon/MDSMonitor: skip offline ERR for down fs

The idea here is that a file system marked down is considered intentionally
archived by the sysadmin. Consider the future use-case where there are dozens
or hundreds of file systems in use by dynamic applications that come and go.  A
file system may only be brought up when the corresponding application is
online.

Fixes: http://tracker.ceph.com/issues/20611
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

diff --git a/src/mds/MDSMap.cc b/src/mds/MDSMap.cc
index 7b012aa21aa2d8acf23b8b194f8ff13aa1519faa..1b277f105011e277d9ea2c9442dc8562b0dcd60c 100644 (file)
--- a/src/mds/MDSMap.cc
+++ b/src/mds/MDSMap.cc
@@ -496,7 +496,7 @@ void MDSMap::get_health_checks(health_check_map_t *checks) const
   }
 
   // MDS_ALL_DOWN
-  if ((mds_rank_t)get_num_up_mds() == 0) {
+  if ((mds_rank_t)get_num_up_mds() == 0 && get_max_mds() > 0) {
     health_check_t &check = checks->add(
       "MDS_ALL_DOWN", HEALTH_ERR,
       "%num% filesystem%plurals% %isorare% offline");