-
Bug
-
Resolution: Unresolved
-
Medium
-
None
-
None
-
None
-
3
-
9223372036854775807
When an MDT restarts while a peer MDT (MDT0000) is serving lookups for objects that live on the restarting MDT, MDT0000's regular mdt service threads all block in osp_remote_sync waiting on the restarting MDT. The restarting MDT's recovery then needs MDT0000 to accept its LWP MDS_CONNECT. That connect arrives with no export, so it is a normal-priority request, and the pool will not give it one of the threads reserved for high-priority work. It never gets serviced, the LWP never reaches FULL, FLD lookups can never be sent, and recovery waits forever.