From: Xuehan Xu <xuxuehan@qianxin.com>
Date: Tue, 15 Oct 2024 05:59:50 +0000 (+0800)
Subject: crimson/osd/pg: trigger PG::wait_for_active_blocker only when the pg is
X-Git-Tag: v20.0.0~718^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=11e3ecabf971a7d1b0518239684a4ec2b76992d3;p=ceph.git

crimson/osd/pg: trigger PG::wait_for_active_blocker only when the pg is
going active

Fixes: https://tracker.ceph.com/issues/68541
Signed-off-by: Xuehan Xu <xuxuehan@qianxin.com>
---

diff --git a/src/crimson/osd/pg.cc b/src/crimson/osd/pg.cc
index 97d48c1fa454..d17577792be9 100644
--- a/src/crimson/osd/pg.cc
+++ b/src/crimson/osd/pg.cc
@@ -389,7 +389,13 @@ void PG::on_replica_activate()
 
 void PG::on_activate_complete()
 {
-  wait_for_active_blocker.unblock();
+  /* Confusingly, on_activate_complete is invoked when the primary and replicas
+   * have recorded the current interval.  At that point, the PG may either become
+   * ACTIVE or PEERED, depending on whether the acting set is eligible for client
+   * IO.  Only unblock wait_for_active_blocker if we actually became ACTIVE */
+  if (peering_state.is_active()) {
+    wait_for_active_blocker.unblock();
+  }
 
   if (peering_state.needs_recovery()) {
     logger().info("{}: requesting recovery",