From: Sage Weil <sage@redhat.com>
Date: Sun, 23 Jul 2017 14:26:47 +0000 (-0400)
Subject: mon/OSDMonitor: remove zeroed new_state updates
X-Git-Tag: v12.1.2~148^2~1
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=a1611c89df519d3ac99e43d018a9524f94de96e2;p=ceph.git

mon/OSDMonitor: remove zeroed new_state updates

These are interpreted as CEPH_OSD_UP in apply_incremental for legacy
reasons--we don't want them!

Fixes: http://tracker.ceph.com/issues/20751
Signed-off-by: Sage Weil <sage@redhat.com>
---

diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index 82fd1cfae43..9ab39128cb2 100644
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -942,6 +942,18 @@ void OSDMonitor::encode_pending(MonitorDBStore::TransactionRef t)
   }
   mapping_job.reset();
 
+  // ensure we don't have blank new_state updates.  these are interrpeted as
+  // CEPH_OSD_UP (and almost certainly not what we want!).
+  auto p = pending_inc.new_state.begin();
+  while (p != pending_inc.new_state.end()) {
+    if (p->second == 0) {
+      dout(10) << "new_state for osd." << p->first << " is 0, removing" << dendl;
+      p = pending_inc.new_state.erase(p);
+    } else {
+      ++p;
+    }
+  }
+
   bufferlist bl;
 
   {