From: Radoslaw Zarzynski <rzarzyns@redhat.com>
Date: Tue, 5 Aug 2025 14:11:59 +0000 (+0200)
Subject: osd: stop scrub_purged_snaps() from ignoring osd_beacon_report_interval
X-Git-Tag: testing/wip-jcollin-testing-20251004.055540-tentacle~23^2
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=814b2766f4e696d38aab9b194ef445e1388e0529;p=ceph-ci.git

osd: stop scrub_purged_snaps() from ignoring osd_beacon_report_interval

OSD beacons could be burdersome to the enitre cluster, as they lead
to generation of new `OSDMap` epochs. Therefore their frequency is
restricted through `osd_beacon_report_interval` to 5 mins by default.

Unfortunately, the `OSD::send_purged_snaps()` is unaware about this
policy with the net result being storm of OSDMaps. This patch unifies
its behavior with `OSD::tick_without_osd_lock()`.

Fixes: https://tracker.ceph.com/issues/72412
Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
(cherry picked from commit d4f90ea6d3c5c565d1ddc6f9ecd9499048f054c0)
---

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index a1175bfc9d7..b354e1d4306 100644
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -6512,20 +6512,7 @@ void OSD::tick_without_osd_lock()
     service.get_scrub_services().initiate_scrub(service.is_recovery_active());
     service.promote_throttle_recalibrate();
     resume_creating_pg();
-    bool need_send_beacon = false;
-    const auto now = ceph::coarse_mono_clock::now();
-    {
-      // borrow lec lock to pretect last_sent_beacon from changing
-      std::lock_guard l{min_last_epoch_clean_lock};
-      const auto elapsed = now - last_sent_beacon;
-      if (std::chrono::duration_cast<std::chrono::seconds>(elapsed).count() >
-        cct->_conf->osd_beacon_report_interval) {
-        need_send_beacon = true;
-      }
-    }
-    if (need_send_beacon) {
-      send_beacon(now);
-    }
+    maybe_send_beacon();
   }
 
   mgrc.update_daemon_health(get_health_metrics());
@@ -7423,6 +7410,26 @@ void OSD::send_beacon(const ceph::coarse_mono_clock::time_point& now)
   }
 }
 
+void OSD::maybe_send_beacon()
+{
+  bool need_send_beacon = false;
+  const auto now = ceph::coarse_mono_clock::now();
+  {
+    // borrow lec lock to protect last_sent_beacon from changing
+    std::lock_guard l{min_last_epoch_clean_lock};
+    const auto elapsed = now - last_sent_beacon;
+    if (std::chrono::duration_cast<std::chrono::seconds>(elapsed).count() >
+      cct->_conf->osd_beacon_report_interval) {
+      need_send_beacon = true;
+    }
+  }
+  if (need_send_beacon) {
+    send_beacon(now);
+  } else {
+    dout(20) << __func__ << " beacon would be too frequent; skipping" << dendl;
+  }
+}
+
 void OSD::handle_command(MCommand *m)
 {
   ConnectionRef con = m->get_connection();
@@ -7509,7 +7516,7 @@ void OSD::scrub_purged_snaps()
   int tr = store->queue_transaction(service.meta_ch, std::move(t), nullptr);
   ceph_assert(tr == 0);
   if (is_active()) {
-    send_beacon(ceph::coarse_mono_clock::now());
+    maybe_send_beacon();
   }
   dout(10) << __func__ << " done" << dendl;
 }
diff --git a/src/osd/OSD.h b/src/osd/OSD.h
index 84f70c86a40..cd2a492240b 100644
--- a/src/osd/OSD.h
+++ b/src/osd/OSD.h
@@ -1882,6 +1882,7 @@ protected:
   // which pgs were scanned for min_lec
   std::vector<pg_t> min_last_epoch_clean_pgs;
   void send_beacon(const ceph::coarse_mono_clock::time_point& now);
+  void maybe_send_beacon();
 
   ceph_tid_t get_tid() {
     return service.get_tid();