From 204f0f0536ace738091569ed3ff05aa28cd1577a Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@redhat.com>
Date: Thu, 30 Nov 2017 09:19:37 -0600
Subject: [PATCH] osd: teach mclock about peering_event

Signed-off-by: Sage Weil <sage@redhat.com>
---
 src/common/legacy_config_opts.h |  3 ++
 src/common/options.cc           | 60 +++++++++++++++++++++++++++++++++
 src/osd/mClockOpClassSupport.cc |  5 ++-
 src/osd/mClockOpClassSupport.h  |  7 +++-
 4 files changed, 73 insertions(+), 2 deletions(-)

diff --git a/src/common/legacy_config_opts.h b/src/common/legacy_config_opts.h
index ec32f3d2556..66b9c31ed12 100644
--- a/src/common/legacy_config_opts.h
+++ b/src/common/legacy_config_opts.h
@@ -670,6 +670,9 @@ OPTION(osd_op_queue_mclock_recov_lim, OPT_DOUBLE)
 OPTION(osd_op_queue_mclock_scrub_res, OPT_DOUBLE)
 OPTION(osd_op_queue_mclock_scrub_wgt, OPT_DOUBLE)
 OPTION(osd_op_queue_mclock_scrub_lim, OPT_DOUBLE)
+OPTION(osd_op_queue_mclock_peering_event_res, OPT_DOUBLE)
+OPTION(osd_op_queue_mclock_peering_event_wgt, OPT_DOUBLE)
+OPTION(osd_op_queue_mclock_peering_event_lim, OPT_DOUBLE)
 
 OPTION(osd_ignore_stale_divergent_priors, OPT_BOOL) // do not assert on divergent_prior entries which aren't in the log and whose on-disk objects are newer
 
diff --git a/src/common/options.cc b/src/common/options.cc
index e75f67b08b7..548e886392f 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -2298,6 +2298,66 @@ std::vector<Option> get_global_options() {
     .add_see_also("osd_op_queue_mclock_scrub_res")
     .add_see_also("osd_op_queue_mclock_scrub_wgt"),
 
+    Option("osd_op_queue_mclock_peering_event_res", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
+    .set_default(0.0)
+    .set_description("mclock reservation of peering events")
+    .set_long_description("mclock reservation of scrub requests when osd_op_queue is either 'mclock_opclass' or 'mclock_client'; higher values increase the reservation")
+    .add_see_also("osd_op_queue")
+    .add_see_also("osd_op_queue_mclock_client_op_res")
+    .add_see_also("osd_op_queue_mclock_client_op_wgt")
+    .add_see_also("osd_op_queue_mclock_client_op_lim")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_res")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_wgt")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_lim")
+    .add_see_also("osd_op_queue_mclock_snap_res")
+    .add_see_also("osd_op_queue_mclock_snap_wgt")
+    .add_see_also("osd_op_queue_mclock_snap_lim")
+    .add_see_also("osd_op_queue_mclock_recov_res")
+    .add_see_also("osd_op_queue_mclock_recov_wgt")
+    .add_see_also("osd_op_queue_mclock_recov_lim")
+    .add_see_also("osd_op_queue_mclock_scrub_wgt")
+    .add_see_also("osd_op_queue_mclock_scrub_lim"),
+
+    Option("osd_op_queue_mclock_peering_event_wgt", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
+    .set_default(1.0)
+    .set_description("mclock weight of peering events")
+    .set_long_description("mclock weight of scrub requests when osd_op_queue is either 'mclock_opclass' or 'mclock_client'; higher values increase the weight")
+    .add_see_also("osd_op_queue")
+    .add_see_also("osd_op_queue_mclock_client_op_res")
+    .add_see_also("osd_op_queue_mclock_client_op_wgt")
+    .add_see_also("osd_op_queue_mclock_client_op_lim")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_res")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_wgt")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_lim")
+    .add_see_also("osd_op_queue_mclock_snap_res")
+    .add_see_also("osd_op_queue_mclock_snap_wgt")
+    .add_see_also("osd_op_queue_mclock_snap_lim")
+    .add_see_also("osd_op_queue_mclock_recov_res")
+    .add_see_also("osd_op_queue_mclock_recov_wgt")
+    .add_see_also("osd_op_queue_mclock_recov_lim")
+    .add_see_also("osd_op_queue_mclock_scrub_res")
+    .add_see_also("osd_op_queue_mclock_scrub_lim"),
+
+    Option("osd_op_queue_mclock_peering_event_lim", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
+    .set_default(0.001)
+    .set_description("mclock weight of limit peering events")
+    .set_long_description("mclock weight of limit requests when osd_op_queue is either 'mclock_opclass' or 'mclock_client'; higher values increase the limit")
+    .add_see_also("osd_op_queue")
+    .add_see_also("osd_op_queue_mclock_client_op_res")
+    .add_see_also("osd_op_queue_mclock_client_op_wgt")
+    .add_see_also("osd_op_queue_mclock_client_op_lim")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_res")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_wgt")
+    .add_see_also("osd_op_queue_mclock_osd_rep_op_lim")
+    .add_see_also("osd_op_queue_mclock_snap_res")
+    .add_see_also("osd_op_queue_mclock_snap_wgt")
+    .add_see_also("osd_op_queue_mclock_snap_lim")
+    .add_see_also("osd_op_queue_mclock_recov_res")
+    .add_see_also("osd_op_queue_mclock_recov_wgt")
+    .add_see_also("osd_op_queue_mclock_recov_lim")
+    .add_see_also("osd_op_queue_mclock_scrub_res")
+    .add_see_also("osd_op_queue_mclock_scrub_wgt"),
+
     Option("osd_ignore_stale_divergent_priors", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
     .set_default(false)
     .set_description(""),
diff --git a/src/osd/mClockOpClassSupport.cc b/src/osd/mClockOpClassSupport.cc
index afbc2f739c5..3849ad9c680 100644
--- a/src/osd/mClockOpClassSupport.cc
+++ b/src/osd/mClockOpClassSupport.cc
@@ -38,7 +38,10 @@ namespace ceph {
 	    cct->_conf->osd_op_queue_mclock_recov_lim),
       scrub(cct->_conf->osd_op_queue_mclock_scrub_res,
 	    cct->_conf->osd_op_queue_mclock_scrub_wgt,
-	    cct->_conf->osd_op_queue_mclock_scrub_lim)
+	    cct->_conf->osd_op_queue_mclock_scrub_lim),
+      peering_event(cct->_conf->osd_op_queue_mclock_peering_event_res,
+		    cct->_conf->osd_op_queue_mclock_peering_event_wgt,
+		    cct->_conf->osd_op_queue_mclock_peering_event_lim)
     {
       constexpr int rep_ops[] = {
 	MSG_OSD_REPOP,
diff --git a/src/osd/mClockOpClassSupport.h b/src/osd/mClockOpClassSupport.h
index 8f57c9b0641..a33e57a5d61 100644
--- a/src/osd/mClockOpClassSupport.h
+++ b/src/osd/mClockOpClassSupport.h
@@ -28,7 +28,7 @@ namespace ceph {
     using op_item_type_t = OpQueueItem::OpQueueable::op_type_t;
     
     enum class osd_op_type_t {
-      client_op, osd_rep_op, bg_snaptrim, bg_recovery, bg_scrub
+      client_op, osd_rep_op, bg_snaptrim, bg_recovery, bg_scrub, peering_event
     };
 
     class OpClassClientInfoMgr {
@@ -37,6 +37,7 @@ namespace ceph {
       crimson::dmclock::ClientInfo snaptrim;
       crimson::dmclock::ClientInfo recov;
       crimson::dmclock::ClientInfo scrub;
+      crimson::dmclock::ClientInfo peering_event;
 
       static constexpr std::size_t rep_op_msg_bitset_size = 128;
       std::bitset<rep_op_msg_bitset_size> rep_op_msg_bitset;
@@ -59,6 +60,8 @@ namespace ceph {
 	  return &recov;
 	case osd_op_type_t::bg_scrub:
 	  return &scrub;
+	case osd_op_type_t::peering_event:
+	  return &peering_event;
 	default:
 	  ceph_abort();
 	  return nullptr;
@@ -77,6 +80,8 @@ namespace ceph {
 	  return osd_op_type_t::bg_recovery;
 	case op_item_type_t::bg_scrub:
 	  return osd_op_type_t::bg_scrub;
+	case op_item_type_t::peering_event:
+	  return osd_op_type_t::peering_event;
 	default:
 	  ceph_abort();
 	}
-- 
2.47.3