From 66690ea3143ac5097f7c7f3118f2d00fed30cc4b Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@redhat.com>
Date: Thu, 5 Dec 2019 12:59:31 -0600
Subject: [PATCH] mgr/DaemonServer: fix 'osd ok-to-stop' for EC pools

We need to pay attention to account for CRUSH_ITEM_NONE entries in the
EC PG acting set.

Fixes: https://tracker.ceph.com/issues/43151
Signed-off-by: Sage Weil <sage@redhat.com>
---
 qa/standalone/misc/ok-to-stop.sh | 48 ++++++++++++++++++++++++++++++++
 src/mgr/DaemonServer.cc          |  8 ++++--
 2 files changed, 54 insertions(+), 2 deletions(-)

diff --git a/qa/standalone/misc/ok-to-stop.sh b/qa/standalone/misc/ok-to-stop.sh
index e53b28d1eb28b..5e4a4180d9c7e 100755
--- a/qa/standalone/misc/ok-to-stop.sh
+++ b/qa/standalone/misc/ok-to-stop.sh
@@ -237,5 +237,53 @@ function TEST_0_mds() {
     kill_daemons $dir KILL mds.a
 }
 
+function TEST_0_osd() {
+    local dir=$1
+
+    CEPH_ARGS="$ORIG_CEPH_ARGS --mon-host=$CEPH_MON_A "
+
+    run_mon $dir a --public-addr=$CEPH_MON_A || return 1
+    run_mgr $dir x || return 1
+    run_osd $dir 0 || return 1
+    run_osd $dir 1 || return 1
+    run_osd $dir 2 || return 1
+    run_osd $dir 3 || return 1
+
+    ceph osd erasure-code-profile set ec-profile m=2 k=2 crush-failure-domain=osd || return 1
+    ceph osd pool create ec erasure ec-profile || return 1
+
+    wait_for_clean || return 1
+
+    # with min_size 3, we can stop only 1 osd
+    ceph osd pool set ec min_size 3 || return 1
+    flush_pg_stats || return 1
+
+    ceph osd ok-to-stop 0 || return 1
+    ceph osd ok-to-stop 1 || return 1
+    ceph osd ok-to-stop 2 || return 1
+    ceph osd ok-to-stop 3 || return 1
+    ! ceph osd ok-to-stop 0 1 || return 1
+    ! ceph osd ok-to-stop 2 3 || return 1
+
+    # with min_size 2 we can stop 1 osds
+    ceph osd pool set ec min_size 2 || return 1
+    flush_pg_stats || return 1
+
+    ceph osd ok-to-stop 0 1 || return 1
+    ceph osd ok-to-stop 2 3 || return 1
+    ! ceph osd ok-to-stop 0 1 2 || return 1
+    ! ceph osd ok-to-stop 1 2 3 || return 1
+
+    # we should get the same result with one of the osds already down
+    kill_daemons $dir TERM osd.0 || return 1
+    ceph osd down 0 || return 1
+    flush_pg_stats || return 1
+
+    ceph osd ok-to-stop 0 || return 1
+    ceph osd ok-to-stop 0 1 || return 1
+    ! ceph osd ok-to-stop 0 1 2 || return 1
+    ! ceph osd ok-to-stop 1 2 3 || return 1
+}
+
 
 main ok-to-stop "$@"
diff --git a/src/mgr/DaemonServer.cc b/src/mgr/DaemonServer.cc
index 7b81ffff02720..2bc739157e039 100644
--- a/src/mgr/DaemonServer.cc
+++ b/src/mgr/DaemonServer.cc
@@ -1588,7 +1588,9 @@ bool DaemonServer::_handle_command(
 		found = true;
 		continue;
 	      }
-	      pg_acting.insert(anm.osd);
+	      if (anm.osd != CRUSH_ITEM_NONE) {
+		pg_acting.insert(anm.osd);
+	      }
 	    }
 	  } else {
 	    for (auto& a : q.second.acting) {
@@ -1596,7 +1598,9 @@ bool DaemonServer::_handle_command(
 		found = true;
 		continue;
 	      }
-	      pg_acting.insert(a);
+	      if (a != CRUSH_ITEM_NONE) {
+		pg_acting.insert(a);
+	      }
 	    }
 	  }
 	  if (!found) {
-- 
2.39.5