From: Sage Weil <sage@redhat.com>
Date: Mon, 29 Apr 2019 21:17:01 +0000 (-0500)
Subject: mgr: fix pgp_num vs merge pacer
X-Git-Tag: v14.2.2~173^2~2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=89e8e2a534f72d9b23b85b151f97ff2b4c27b9b1;p=ceph.git

mgr: fix pgp_num vs merge pacer

This condition should only kick in for pgp_num reductions, or else we'll
end up jump to the end instead of slowing down merges.  (For splits, we
don't need to pace like this at all.)

Fixes 76503a1438fa1f166d2c230c73ca8d7b67e6468d

Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit e6113c172523906ee2f5a317ddb358e215d0a448)
---

diff --git a/src/mgr/DaemonServer.cc b/src/mgr/DaemonServer.cc
index 38143ecc227..1a248eebf33 100644
--- a/src/mgr/DaemonServer.cc
+++ b/src/mgr/DaemonServer.cc
@@ -2601,7 +2601,8 @@ void DaemonServer::adjust_pgs()
 	      }
 	      dout(20) << " room " << room << " estmax " << estmax
 		       << " delta " << delta << " next " << next << dendl;
-	      if (p.get_pgp_num_target() == p.get_pg_num_target()) {
+	      if (p.get_pgp_num_target() == p.get_pg_num_target() &&
+		  p.get_pgp_num_target() < p.get_pg_num()) {
 		// since pgp_num is tracking pg_num, ceph is handling
 		// pgp_num.  so, be responsible: don't let pgp_num get
 		// too far out ahead of merges (if we are merging).