OSD: add recovery_wq suicide timeout

Signed-off-by: Samuel Just <sjust@redhat.com>
(cherry picked from commit 85311b656852af75bfbbc6699f92fc6aa233c316)

Conflicts:
	src/common/config_opts.h
           place OPTION(osd_command_thread_suicide_timeout, ...)
           directly after OPTION(osd_command_thread_timeout, ...)

diff --git a/src/common/config_opts.h b/src/common/config_opts.h
index 5ca307102d5a92eece2ef517b0ade1000019376f..de01d636bc05f82788de83bd398d43f1010b5e0b 100644 (file)
--- a/src/common/config_opts.h
+++ b/src/common/config_opts.h
@@ -477,6 +477,7 @@ OPTION(osd_backfill_scan_min, OPT_INT, 64)
 OPTION(osd_backfill_scan_max, OPT_INT, 512)
 OPTION(osd_op_thread_timeout, OPT_INT, 15)
 OPTION(osd_recovery_thread_timeout, OPT_INT, 30)
+OPTION(osd_recovery_thread_suicide_timeout, OPT_INT, 300)
 OPTION(osd_snap_trim_thread_timeout, OPT_INT, 60*60*1)
 OPTION(osd_snap_trim_sleep, OPT_FLOAT, 0)
 OPTION(osd_scrub_thread_timeout, OPT_INT, 60)

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 6fbc7d89997bad099eb93e906d3fa77c94a1abe9..3b74e4cdc30d89fc15a57d37f57db5f87834d877 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -959,7 +959,11 @@ OSD::OSD(CephContext *cct_, ObjectStore *store_,
     cct->_conf->osd_command_thread_suicide_timeout,
     &command_tp),
   recovery_ops_active(0),
-  recovery_wq(this, cct->_conf->osd_recovery_thread_timeout, &recovery_tp),
+  recovery_wq(
+    this,
+    cct->_conf->osd_recovery_thread_timeout,
+    cct->_conf->osd_recovery_thread_suicide_timeout,
+    &recovery_tp),
   replay_queue_lock("OSD::replay_queue_lock"),
   snap_trim_wq(this, cct->_conf->osd_snap_trim_thread_timeout, &disk_tp),
   scrub_wq(this, cct->_conf->osd_scrub_thread_timeout, &disk_tp),

diff --git a/src/osd/OSD.h b/src/osd/OSD.h
index 26474e4cd09bde2cc511edf04301beb33e3b7326..0ec34771664059d1e3245f5df4e2c20a170bd18a 100644 (file)
--- a/src/osd/OSD.h
+++ b/src/osd/OSD.h
@@ -1653,8 +1653,8 @@ protected:
 
   struct RecoveryWQ : public ThreadPool::WorkQueue<PG> {
     OSD *osd;
-    RecoveryWQ(OSD *o, time_t ti, ThreadPool *tp)
-      : ThreadPool::WorkQueue<PG>("OSD::RecoveryWQ", ti, ti*10, tp), osd(o) {}
+    RecoveryWQ(OSD *o, time_t ti, time_t si, ThreadPool *tp)
+      : ThreadPool::WorkQueue<PG>("OSD::RecoveryWQ", ti, si, tp), osd(o) {}
 
     bool _empty() {
       return osd->recovery_queue.empty();