osd, mon: Add new pg states recovery_unfound and backfill_unfound

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 7f8b0ce9e681f727d8217e3ed74a1a3355f364f3)

Conflicts:
	src/osd/PG.h (trivial)
	src/osd/osd_types.h (trivial)

diff --git a/src/mon/PGMap.cc b/src/mon/PGMap.cc
index 6fcc3c0721afba4ba79cfc83ea15d6e2cffaeeab..36e1d8dee688eae4af77828c605d50ef8e37f063 100644 (file)
--- a/src/mon/PGMap.cc
+++ b/src/mon/PGMap.cc
@@ -2663,6 +2663,8 @@ void PGMap::get_health_checks(
     { PG_STATE_INCOMPLETE,       {UNAVAILABLE, {}} },
     { PG_STATE_REPAIR,           {DAMAGED,     {}} },
     { PG_STATE_SNAPTRIM_ERROR,   {DAMAGED,     {}} },
+    { PG_STATE_RECOVERY_UNFOUND, {DAMAGED,     {}} },
+    { PG_STATE_BACKFILL_UNFOUND, {DAMAGED,     {}} },
     { PG_STATE_BACKFILL_TOOFULL, {DEGRADED_FULL, {}} },
     { PG_STATE_RECOVERY_TOOFULL, {DEGRADED_FULL, {}} },
     { PG_STATE_DEGRADED,         {DEGRADED,    {}} },

diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index 3a919cdbab8a1b45a4f3ff032ad9a133ca15ba7e..b3dd2d9be98d01433ee58f2f7a3f7b31f68ca366 100644 (file)
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -6440,6 +6440,7 @@ PG::RecoveryState::Backfilling::react(const UnfoundBackfill &c)
   ldout(pg->cct, 10) << "backfill has unfound, can't continue" << dendl;
   pg->osd->local_reserver.cancel_reservation(pg->info.pgid);
 
+  pg->state_set(PG_STATE_BACKFILL_UNFOUND);
   pg->state_clear(PG_STATE_BACKFILLING);
 
   for (set<pg_shard_t>::iterator it = pg->backfill_targets.begin();
@@ -6644,7 +6645,7 @@ void PG::RecoveryState::NotBackfilling::exit()
 {
   context< RecoveryMachine >().log_exit(state_name, enter_time);
   PG *pg = context< RecoveryMachine >().pg;
-  pg->state_clear(PG_STATE_UNFOUND);
+  pg->state_clear(PG_STATE_BACKFILL_UNFOUND);
   utime_t dur = ceph_clock_now() - enter_time;
   pg->osd->recoverystate_perf->tinc(rs_notbackfilling_latency, dur);
 }
@@ -6663,7 +6664,7 @@ void PG::RecoveryState::NotRecovering::exit()
 {
   context< RecoveryMachine >().log_exit(state_name, enter_time);
   PG *pg = context< RecoveryMachine >().pg;
-  pg->state_clear(PG_STATE_UNFOUND);
+  pg->state_clear(PG_STATE_RECOVERY_UNFOUND);
   utime_t dur = ceph_clock_now() - enter_time;
   pg->osd->recoverystate_perf->tinc(rs_notrecovering_latency, dur);
 }
@@ -7045,6 +7046,7 @@ PG::RecoveryState::Recovering::react(const UnfoundRecovery &evt)
 {
   PG *pg = context< RecoveryMachine >().pg;
   ldout(pg->cct, 10) << "recovery has unfound, can't continue" << dendl;
+  pg->state_set(PG_STATE_RECOVERY_UNFOUND);
   pg->state_clear(PG_STATE_RECOVERING);
   pg->osd->local_reserver.cancel_reservation(pg->info.pgid);
   release_reservations(true);

diff --git a/src/osd/PG.h b/src/osd/PG.h
index 8adb0638994047be581f45f26f9513e11f093a1f..ccf6d88f90989db1235fefbf51b7f127e0daad6c 100644 (file)
--- a/src/osd/PG.h
+++ b/src/osd/PG.h
@@ -2384,6 +2384,8 @@ protected:
   bool       is_activating() const { return state_test(PG_STATE_ACTIVATING); }
   bool       is_peering() const { return state_test(PG_STATE_PEERING); }
   bool       is_down() const { return state_test(PG_STATE_DOWN); }
+  bool       is_recovery_unfound() const { return state_test(PG_STATE_RECOVERY_UNFOUND); }
+  bool       is_backfill_unfound() const { return state_test(PG_STATE_BACKFILL_UNFOUND); }
   bool       is_incomplete() const { return state_test(PG_STATE_INCOMPLETE); }
   bool       is_clean() const { return state_test(PG_STATE_CLEAN); }
   bool       is_degraded() const { return state_test(PG_STATE_DEGRADED); }

diff --git a/src/osd/osd_types.cc b/src/osd/osd_types.cc
index b22001af6f003d62a0d4cdecf30b6ac55f47b382..102844ca7beb03ad344c97eb75ef6d176b976df7 100644 (file)
--- a/src/osd/osd_types.cc
+++ b/src/osd/osd_types.cc
@@ -811,6 +811,10 @@ std::string pg_state_string(int state)
     oss << "forced_recovery+";
   if (state & PG_STATE_DOWN)
     oss << "down+";
+  if (state & PG_STATE_RECOVERY_UNFOUND)
+    oss << "recovery_unfound+";
+  if (state & PG_STATE_BACKFILL_UNFOUND)
+    oss << "backfill_unfound+";
   if (state & PG_STATE_UNDERSIZED)
     oss << "undersized+";
   if (state & PG_STATE_DEGRADED)
@@ -862,6 +866,10 @@ boost::optional<uint64_t> pg_string_state(const std::string& state)
     type = PG_STATE_CLEAN;
   else if (state == "down")
     type = PG_STATE_DOWN;
+  else if (state == "recovery_unfound")
+    type = PG_STATE_RECOVERY_UNFOUND;
+  else if (state == "backfill_unfound")
+    type = PG_STATE_BACKFILL_UNFOUND;
   else if (state == "scrubbing")
     type = PG_STATE_SCRUBBING;
   else if (state == "degraded")

diff --git a/src/osd/osd_types.h b/src/osd/osd_types.h
index a820c8f6bfa70e8e63881310e5b038ca3fee868b..6fdb6d6d41186b640aff27d91520b2884f993006 100644 (file)
--- a/src/osd/osd_types.h
+++ b/src/osd/osd_types.h
@@ -971,8 +971,8 @@ inline ostream& operator<<(ostream& out, const osd_stat_t& s) {
 #define PG_STATE_ACTIVE       (1<<1)  // i am active.  (primary: replicas too)
 #define PG_STATE_CLEAN        (1<<2)  // peers are complete, clean of stray replicas.
 #define PG_STATE_DOWN         (1<<4)  // a needed replica is down, PG offline
-//#define PG_STATE_REPLAY       (1<<5)  // crashed, waiting for replay
-//#define PG_STATE_STRAY      (1<<6)  // i must notify the primary i exist.
+#define PG_STATE_RECOVERY_UNFOUND   (1<<5)  // recovery stopped due to unfound
+#define PG_STATE_BACKFILL_UNFOUND   (1<<6)  // backfill stopped due to unfound
 //#define PG_STATE_SPLITTING    (1<<7)  // i am splitting
 #define PG_STATE_SCRUBBING    (1<<8)  // scrubbing
 //#define PG_STATE_SCRUBQ       (1<<9)  // queued for scrub