osd: fix requeue order of dup ops

The waiting_for_ondisk (and ack) maps get dups of ops that are in progress.
If we have a peering change in which the role does not change, we will
requeue the in-progress ops but leave these in the waiting_for_ondisk
maps, which will then trigger an assert the next time we examine that map
and find it didn't match up with what we expected.

Fix this by requeuing these on any peering reset in on_change().  This
keeps the two queues in sync.

Fixes: #2956
Signed-off-by: Sage Weil <sage@inktank.com>

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index 50cde51a948df36192d0df7fcf702721afa082c9..296bffb4387b560e7e8570a87f2bb7faebaf3308 100644 (file)
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -5793,6 +5793,15 @@ void ReplicatedPG::on_change()
   requeue_ops(waiting_for_all_missing);
   waiting_for_all_missing.clear();
 
+  // take commit waiters; these are dups of what
+  // apply_and_flush_repops() will requeue.
+  for (map<eversion_t, list<OpRequestRef> >::iterator p = waiting_for_ondisk.begin();
+       p != waiting_for_ondisk.end();
+       p++)
+    requeue_ops(p->second);
+  waiting_for_ondisk.clear();
+  waiting_for_ack.clear();
+
   // this will requeue ops we were working on but didn't finish
   apply_and_flush_repops(is_primary());
 
@@ -5808,18 +5817,9 @@ void ReplicatedPG::on_change()
 void ReplicatedPG::on_role_change()
 {
   dout(10) << "on_role_change" << dendl;
-
-  // take commit waiters
-  for (map<eversion_t, list<OpRequestRef> >::iterator p = waiting_for_ondisk.begin();
-       p != waiting_for_ondisk.end();
-       p++)
-    requeue_ops(p->second);
-  waiting_for_ondisk.clear();
-  waiting_for_ack.clear();
 }
 
 
-
 // clear state.  called on recovery completion AND cancellation.
 void ReplicatedPG::_clear_recovery_state()
 {