From: Xiubo Li <xiubli@redhat.com>
Date: Tue, 22 Feb 2022 07:03:35 +0000 (+0800)
Subject: mds: add mds_heartbeat_reset_grace option support
X-Git-Tag: v16.2.11~540^2~2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=3da4368aaa0ec0c0539002a1d0f32688557f14c6;p=ceph.git

mds: add mds_heartbeat_reset_grace option support

Signed-off-by: Xiubo Li <xiubli@redhat.com>
(cherry picked from commit 109e370c5f4f18d02386086b8babd7ef6a5c578a)

Conflicts:
	src/common/options/mds.yaml.in
---

diff --git a/src/common/options.cc b/src/common/options.cc
index 7fda07387233..06d300b5c2bd 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -8344,6 +8344,11 @@ std::vector<Option> get_mds_options() {
     .set_default(15)
     .set_description("tolerance in seconds for MDS internal heartbeat"),
 
+    Option("mds_heartbeat_reset_grace", Option::TYPE_UINT, Option::LEVEL_ADVANCED)
+    .set_default(1000)
+    .set_description("the basic unit of tolerance in how many circles in a loop, which will"
+	"keep running by holding the mds_lock, it must trigger to reset heartbeat"),
+
     Option("mds_enforce_unique_name", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
     .set_default(true)
     .set_description("require MDS name is unique in the cluster"),
diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
index b7ff6c9c7b89..de411b24cb60 100644
--- a/src/mds/MDSRank.cc
+++ b/src/mds/MDSRank.cc
@@ -538,6 +538,7 @@ MDSRank::MDSRank(
   server = new Server(this, &metrics_handler);
   locker = new Locker(this, mdcache);
 
+  _heartbeat_reset_grace = g_conf().get_val<uint64_t>("mds_heartbeat_reset_grace");
   heartbeat_grace = g_conf().get_val<double>("mds_heartbeat_grace");
   op_tracker.set_complaint_and_threshold(cct->_conf->mds_op_complaint_time,
                                          cct->_conf->mds_op_log_threshold);
@@ -3749,6 +3750,7 @@ const char** MDSRankDispatcher::get_tracked_conf_keys() const
     "mds_recall_warning_decay_rate",
     "mds_request_load_average_decay_rate",
     "mds_session_cache_liveness_decay_rate",
+    "mds_heartbeat_reset_grace",
     "mds_heartbeat_grace",
     "mds_session_cap_acquisition_decay_rate",
     "mds_max_caps_per_client",
@@ -3766,6 +3768,9 @@ void MDSRankDispatcher::handle_conf_change(const ConfigProxy& conf, const std::s
 {
   // XXX with or without mds_lock!
 
+  if (changed.count("mds_heartbeat_reset_grace")) {
+    _heartbeat_reset_grace = conf.get_val<uint64_t>("mds_heartbeat_reset_grace");
+  }
   if (changed.count("mds_heartbeat_grace")) {
     heartbeat_grace = conf.get_val<double>("mds_heartbeat_grace");
   }
diff --git a/src/mds/MDSRank.h b/src/mds/MDSRank.h
index 7fe06420e8e7..3515ee5367b4 100644
--- a/src/mds/MDSRank.h
+++ b/src/mds/MDSRank.h
@@ -268,6 +268,9 @@ class MDSRank {
      * of code while holding the mds_lock
      */
     void heartbeat_reset();
+    int heartbeat_reset_grace(int count=1) {
+      return count * _heartbeat_reset_grace;
+    }
 
     /**
      * Report state DAMAGED to the mon, and then pass on to respawn().  Call
@@ -578,6 +581,7 @@ class MDSRank {
 
     ceph::heartbeat_handle_d *hb = nullptr;  // Heartbeat for threads using mds_lock
     double heartbeat_grace;
+    int _heartbeat_reset_grace;
 
     map<mds_rank_t, version_t> peer_mdsmap_epoch;