osd: don't clobber racing pg_info on active pg

Wido saw a pg go active, but an activate log+info update crossed paths with
a pg_notify info, and the primary overwrote it's updated shiny new info
with the stale old info from the replica.  Don't do that.  It causes
problems down the line.  In this case, we got

osd/OSD.cc: In function 'void OSD::generate_backlog(PG*)':
osd/OSD.cc:3863: FAILED assert(!pg->is_active())
 1: (ThreadPool::worker()+0x28f) [0x5b08ff]
 2: (ThreadPool::WorkThread::entry()+0xd) [0x4edb8d]
 3: (Thread::_entry_func(void*)+0xa) [0x46892a]
 4: (()+0x69ca) [0x7f889ff249ca]
 5: (clone()+0x6d) [0x7f889f1446cd]

on the replica because it was active but the primary was restarting peering
due to the bad info.

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 1e7427692106aab7662635f3dede7f658f228a5b..62f98064deb3ee08b482a885543d5875f2f34e2c 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -3239,7 +3239,12 @@ void OSD::handle_pg_notify(MOSDPGNotify *m)
 
     if (pg->peer_info.count(from) &&
        pg->peer_info[from].last_update == it->last_update) {
-      dout(10) << *pg << " got dup osd" << from << " info " << *it << dendl;
+      dout(10) << *pg << " got dup osd" << from << " info " << *it << ", identical to ours" << dendl;
+    } else if (pg->peer_info.count(from) &&
+              pg->is_active()) {
+      dout(10) << *pg << " got dup osd" << from << " info " << *it
+              << " but pg is active, keeping our info " << pg->peer_info[from]
+              << dendl;
     } else {
       dout(10) << *pg << " got osd" << from << " info " << *it << dendl;
       pg->peer_info[from] = *it;