osd/scrub: restart snap trimming after a failed scrub

A followup to PR#45640.
In PR#45640 snap trimming was restarted (if blocked) after all
successful scrubs, and after most scrub failures. Still, a few
failure scenarios did not handle snaptrim restart correctly.

The current PR cleans up and fixes the interaction between
scrub initiation/termination (for whatever cause) and snap
trimming.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
(cherry picked from commit 290e744a9b6c64f3da805056625b963f0eedaf33)

diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index d3345319acbc6cfa117c557d4dfdde53e6c2c160..d77cef0d7b5c552b1156a2124b56dcea82c65a7d 100644 (file)
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -441,7 +441,6 @@ void PG::queue_scrub_after_repair()
   m_scrubber->set_op_parameters(m_planned_scrub);
   dout(15) << __func__ << ": queueing" << dendl;
 
-  m_scrubber->set_queued_or_active();
   osd->queue_scrub_after_repair(this, Scrub::scrub_prio_t::high_priority);
 }
 
@@ -1371,7 +1370,6 @@ Scrub::schedule_result_t PG::sched_scrub()
   m_scrubber->set_op_parameters(m_planned_scrub);
 
   dout(10) << __func__ << ": queueing" << dendl;
-  m_scrubber->set_queued_or_active();
   osd->queue_for_scrub(this, Scrub::scrub_prio_t::low_priority);
   return Scrub::schedule_result_t::scrub_initiated;
 }

diff --git a/src/osd/scrubber/pg_scrubber.cc b/src/osd/scrubber/pg_scrubber.cc
index 5b4ed18b8f494849515e7ba995f69ea679a111d4..25e4a83d9e8b4974f0315bcfbdd9a38bf7f6a838 100644 (file)
--- a/src/osd/scrubber/pg_scrubber.cc
+++ b/src/osd/scrubber/pg_scrubber.cc
@@ -188,9 +188,7 @@ void PgScrubber::initiate_regular_scrub(epoch_t epoch_queued)
     m_fsm->process_event(StartScrub{});
     dout(10) << "scrubber event --<< StartScrub" << dendl;
   } else {
-    clear_queued_or_active();
-    // and just in case snap trimming was blocked by the aborted scrub
-    m_pg->snap_trimmer_scrub_complete();
+    clear_queued_or_active();  // also restarts snap trimming
   }
 }
 
@@ -204,9 +202,7 @@ void PgScrubber::initiate_scrub_after_repair(epoch_t epoch_queued)
     m_fsm->process_event(AfterRepairScrub{});
     dout(10) << "scrubber event --<< AfterRepairScrub" << dendl;
   } else {
-    clear_queued_or_active();
-    // and just in case snap trimming was blocked by the aborted scrub
-    m_pg->snap_trimmer_scrub_complete();
+    clear_queued_or_active();  // also restarts snap trimming
   }
 }
 
@@ -1335,6 +1331,8 @@ void PgScrubber::set_op_parameters(requested_scrub_t& request)
 {
   dout(10) << __func__ << " input: " << request << dendl;
 
+  set_queued_or_active(); // we are fully committed now.
+
   // write down the epoch of starting a new scrub. Will be used
   // to discard stale messages from previous aborted scrubs.
   m_epoch_start = m_pg->get_osdmap_epoch();
@@ -1715,7 +1713,11 @@ void PgScrubber::set_queued_or_active()
 
 void PgScrubber::clear_queued_or_active()
 {
-  m_queued_or_active = false;
+  if (m_queued_or_active) {
+    m_queued_or_active = false;
+    // and just in case snap trimming was blocked by the aborted scrub
+    m_pg->snap_trimmer_scrub_complete();
+  }
 }
 
 bool PgScrubber::is_queued_or_active() const
@@ -1921,9 +1923,6 @@ void PgScrubber::scrub_finish()
   if (m_pg->is_active() && m_pg->is_primary()) {
     m_pg->recovery_state.share_pg_info();
   }
-
-  // we may have blocked the snap trimmer
-  m_pg->snap_trimmer_scrub_complete();
 }
 
 void PgScrubber::on_digest_updates()

diff --git a/src/osd/scrubber/pg_scrubber.h b/src/osd/scrubber/pg_scrubber.h
index e223ae213ab753fc2232657b6c7fdce20c25eedc..eeba65bab79602aa9807ff3156bcd06735d5d94a 100644 (file)
--- a/src/osd/scrubber/pg_scrubber.h
+++ b/src/osd/scrubber/pg_scrubber.h
@@ -429,6 +429,7 @@ class PgScrubber : public ScrubPgIF, public ScrubMachineListener {
   [[nodiscard]] bool was_epoch_changed() const final;
 
   void set_queued_or_active() final;
+  /// Clears `m_queued_or_active` and restarts snaptrimming
   void clear_queued_or_active() final;
 
   void mark_local_map_ready() final;

diff --git a/src/osd/scrubber_common.h b/src/osd/scrubber_common.h
index 2b4d370be11ddb589eb150f6dc98a3aafdd78e2c..078eb60afb9879a59f59cad24992e85b3c3326e0 100644 (file)
--- a/src/osd/scrubber_common.h
+++ b/src/osd/scrubber_common.h
@@ -191,6 +191,8 @@ struct ScrubPgIF {
   /**
    * Manipulate the 'scrubbing request has been queued, or - we are
    * actually scrubbing' Scrubber's flag
+   *
+   * clear_queued_or_active() will also restart any blocked snaptrimming.
    */
   virtual void set_queued_or_active() = 0;
   virtual void clear_queued_or_active() = 0;