OSD: add command_wq suicide timeout

Signed-off-by: Samuel Just <sjust@redhat.com>

diff --git a/src/common/config_opts.h b/src/common/config_opts.h
index 7c4ca61e8d004e3892e2ecbeb94a985281f18526..1183df72e3e56ecb018513503bf0011a3c81e565 100644 (file)
--- a/src/common/config_opts.h
+++ b/src/common/config_opts.h
@@ -592,6 +592,7 @@ OPTION(osd_scrub_finalize_thread_timeout, OPT_INT, 60*10)
 OPTION(osd_scrub_invalid_stats, OPT_BOOL, true)
 OPTION(osd_remove_thread_timeout, OPT_INT, 60*60)
 OPTION(osd_command_thread_timeout, OPT_INT, 10*60)
+OPTION(osd_command_thread_suicide_timeout, OPT_INT, 15*60)
 OPTION(osd_heartbeat_addr, OPT_ADDR, entity_addr_t())
 OPTION(osd_heartbeat_interval, OPT_INT, 6)       // (seconds) how often we ping peers
 OPTION(osd_heartbeat_grace, OPT_INT, 20)         // (seconds) how long before we decide a peer has failed

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 88b08e556055b97dd00bdc6a2e676a55750be13e..b6e04023a9935c788f9e1578fe137bdfd24775e5 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -1515,7 +1515,11 @@ OSD::OSD(CephContext *cct_, ObjectStore *store_,
   pg_stat_queue_lock("OSD::pg_stat_queue_lock"),
   osd_stat_updated(false),
   pg_stat_tid(0), pg_stat_tid_flushed(0),
-  command_wq(this, cct->_conf->osd_command_thread_timeout, &command_tp),
+  command_wq(
+    this,
+    cct->_conf->osd_command_thread_timeout,
+    cct->_conf->osd_command_thread_suicide_timeout,
+    &command_tp),
   recovery_ops_active(0),
   recovery_wq(this, cct->_conf->osd_recovery_thread_timeout, &recovery_tp),
   replay_queue_lock("OSD::replay_queue_lock"),

diff --git a/src/osd/OSD.h b/src/osd/OSD.h
index e183544b487b70c92dee8dc325e5345780cbe84f..b90a0c9cb31863e2b1e63296646245f8c58c215e 100644 (file)
--- a/src/osd/OSD.h
+++ b/src/osd/OSD.h
@@ -1969,8 +1969,8 @@ protected:
   list<Command*> command_queue;
   struct CommandWQ : public ThreadPool::WorkQueue<Command> {
     OSD *osd;
-    CommandWQ(OSD *o, time_t ti, ThreadPool *tp)
-      : ThreadPool::WorkQueue<Command>("OSD::CommandWQ", ti, 0, tp), osd(o) {}
+    CommandWQ(OSD *o, time_t ti, time_t si, ThreadPool *tp)
+      : ThreadPool::WorkQueue<Command>("OSD::CommandWQ", ti, si, tp), osd(o) {}
 
     bool _empty() {
       return osd->command_queue.empty();