mds: defer sending resolves until mdsmap.failed.empty()

There is no point sending resolves while there are still failed nodes,
since we can't complete.  We also trigger an assert if we try to send to
a failed node.  Instead just wait until failed.empty() and then start.

Signed-off-by: Sage Weil <sage.weil@dreamhost.com>

diff --git a/src/mds/MDS.cc b/src/mds/MDS.cc
index efce6e6b61f42d52c938124884dd4cee9c1e42d0..0a1d2f7b0ea153dea7823da57de9a50272660a43 100644 (file)
--- a/src/mds/MDS.cc
+++ b/src/mds/MDS.cc
@@ -958,7 +958,8 @@ void MDS::handle_mds_map(MMDSMap *m)
     if (oldresolve != resolve) {
       dout(10) << " resolve set is " << resolve << ", was " << oldresolve << dendl;
       calc_recovery_set();
-      mdcache->send_resolves();
+      if (!mdsmap->is_any_failed())
+       mdcache->send_resolves();
     }
   }
   

diff --git a/src/mds/MDSMap.h b/src/mds/MDSMap.h
index bc1015f2712b2f584e534f01590f3f3ee7392657..7bcdd272bb093eaae5a53b2ade261fd28d3f5363 100644 (file)
--- a/src/mds/MDSMap.h
+++ b/src/mds/MDSMap.h
@@ -410,6 +410,9 @@ public:
       get_num_mds(STATE_REJOIN) + 
       failed.size();
   }
+  bool is_any_failed() {
+    return failed.size();
+  }
   bool is_rejoining() {  
     // nodes are rejoining cache state
     return