osd: make osd replay interval a per-pool property

Change the config value to only control the interval set when the data
pool is first created (presumably during mkfs).  Start replay interval
based on the pool property.

Signed-off-by: Sage Weil <sage@newdream.net>

diff --git a/src/common/config_opts.h b/src/common/config_opts.h
index 250cdb44e4850c7232b52503debaa55aabfa769a..ab526415982074a079a6544e4d528eefb333bfdf 100644 (file)
--- a/src/common/config_opts.h
+++ b/src/common/config_opts.h
@@ -262,7 +262,7 @@ OPTION(osd_mon_report_interval_max, OPT_INT, 120)
 OPTION(osd_mon_report_interval_min, OPT_INT, 5)  // pg stats, failures, up_thru, boot.
 OPTION(osd_min_down_reporters, OPT_INT, 1)   // number of OSDs who need to report a down OSD for it to count
 OPTION(osd_min_down_reports, OPT_INT, 3)     // number of times a down OSD must be reported for it to count
-OPTION(osd_replay_window, OPT_INT, 45)
+OPTION(osd_default_data_pool_replay_window, OPT_INT, 45)
 OPTION(osd_preserve_trimmed_log, OPT_BOOL, true)
 OPTION(osd_auto_mark_unfound_lost, OPT_BOOL, false)
 OPTION(osd_recovery_delay_start, OPT_FLOAT, 15)

diff --git a/src/osd/OSDMap.cc b/src/osd/OSDMap.cc
index 6ce35f3c366ad4923296b5892e5f59446a60135e..1a3020aa1a4ea5c0760d83152704736abd4db0a1 100644 (file)
--- a/src/osd/OSDMap.cc
+++ b/src/osd/OSDMap.cc
@@ -906,6 +906,8 @@ void OSDMap::build_simple(CephContext *cct, epoch_t e, ceph_fsid_t &fsid,
     pools[pool].lpg_num = lpg_bits ? (1 << (lpg_bits-1)) : 0;
     pools[pool].lpgp_num = lpg_bits ? (1 << (lpg_bits-1)) : 0;
     pools[pool].last_change = epoch;
+    if (p->first == CEPH_DATA_RULE)
+      pools[pool].crash_replay_interval = cct->_conf->osd_default_data_pool_replay_window;
     pool_name[pool] = p->second;
   }
 

diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index eed29b966cf61f4ec8f27983d4287492fa86eb79..47714e5c8968667170c20203b4375ae2698152a8 100644 (file)
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -1331,20 +1331,16 @@ void PG::activate(ObjectStore::Transaction& t, list<Context*>& tfin,
   assert(!is_active());
 
   // -- crash recovery?
-  if (may_need_replay(osd->osdmap)) {
-    if (g_conf->osd_replay_window > 0) {
-      replay_until = ceph_clock_now(g_ceph_context);
-      replay_until += g_conf->osd_replay_window;
-      dout(10) << "crashed, allowing op replay for " << g_conf->osd_replay_window
-              << " until " << replay_until << dendl;
-      state_set(PG_STATE_REPLAY);
-      osd->replay_queue_lock.Lock();
-      osd->replay_queue.push_back(pair<pg_t,utime_t>(info.pgid, replay_until));
-      osd->replay_queue_lock.Unlock();
-    } else {
-      dout(10) << "crashed, but osd_replay_window=0.  skipping replay." << dendl;
-      state_clear(PG_STATE_REPLAY);
-    }
+  if (pool->info.crash_replay_interval > 0 &&
+      may_need_replay(osd->osdmap)) {
+    replay_until = ceph_clock_now(g_ceph_context);
+    replay_until += pool->info.crash_replay_interval;
+    dout(10) << "activate starting replay interval for " << pool->info.crash_replay_interval
+            << " until " << replay_until << dendl;
+    state_set(PG_STATE_REPLAY);
+    osd->replay_queue_lock.Lock();
+    osd->replay_queue.push_back(pair<pg_t,utime_t>(info.pgid, replay_until));
+    osd->replay_queue_lock.Unlock();
   }
 
   // twiddle pg state