mon/PGMap: use new-style options

Signed-off-by: Kefu Chai <kchai@redhat.com>

diff --git a/src/common/legacy_config_opts.h b/src/common/legacy_config_opts.h
index 0887a0d9494a45adc6fc6a1dadc19e8c4de04b32..35dff79e039725f44aa64a6f939bdcdc40428413 100644 (file)
--- a/src/common/legacy_config_opts.h
+++ b/src/common/legacy_config_opts.h
@@ -223,7 +223,6 @@ OPTION(mon_osd_prime_pg_temp, OPT_BOOL)  // prime osdmap with pg mapping changes
 OPTION(mon_osd_prime_pg_temp_max_time, OPT_FLOAT)  // max time to spend priming
 OPTION(mon_osd_prime_pg_temp_max_estimate, OPT_FLOAT) // max estimate of pg total before we do all pgs in parallel
 OPTION(mon_osd_pool_ec_fast_read, OPT_BOOL) // whether turn on fast read on the pool or not
-OPTION(mon_stat_smooth_intervals, OPT_INT)  // smooth stats over last N PGMap maps
 OPTION(mon_election_timeout, OPT_FLOAT)  // on election proposer, max waiting time for all ACKs
 OPTION(mon_lease, OPT_FLOAT)       // lease interval
 OPTION(mon_lease_renew_interval_factor, OPT_FLOAT) // on leader, to renew the lease
@@ -234,11 +233,7 @@ OPTION(mon_clock_drift_allowed, OPT_FLOAT) // allowed clock drift between monito
 OPTION(mon_clock_drift_warn_backoff, OPT_FLOAT) // exponential backoff for clock drift warnings
 OPTION(mon_timecheck_interval, OPT_FLOAT) // on leader, timecheck (clock drift check) interval (seconds)
 OPTION(mon_timecheck_skew_interval, OPT_FLOAT) // on leader, timecheck (clock drift check) interval when in presence of a skew (seconds)
-OPTION(mon_pg_stuck_threshold, OPT_INT) // number of seconds after which pgs can be considered stuck inactive, unclean, etc (see doc/control.rst under dump_stuck for more info)
 OPTION(mon_pg_min_inactive, OPT_U64) // the number of PGs which have to be inactive longer than 'mon_pg_stuck_threshold' before health goes into ERR. 0 means disabled, never go into ERR.
-OPTION(mon_pg_warn_max_object_skew, OPT_FLOAT) // max skew few average in objects per pg
-OPTION(mon_pg_warn_min_objects, OPT_INT)  // do not warn below this object #
-OPTION(mon_pg_warn_min_pool_objects, OPT_INT)  // do not warn on pools below this object #
 OPTION(mon_pg_check_down_all_threshold, OPT_FLOAT) // threshold of down osds after which we check all pgs
 OPTION(mon_cache_target_full_warn_ratio, OPT_FLOAT) // position between pool cache_target_full and max where we start warning
 OPTION(mon_osd_full_ratio, OPT_FLOAT) // what % full makes an OSD "full"
@@ -273,7 +268,6 @@ OPTION(mon_health_to_clog, OPT_BOOL)
 OPTION(mon_health_to_clog_interval, OPT_INT)
 OPTION(mon_health_to_clog_tick_interval, OPT_DOUBLE)
 OPTION(mon_health_preluminous_compat, OPT_BOOL)
-OPTION(mon_health_max_detail, OPT_INT) // max detailed pgs to report in health detail
 OPTION(mon_data_avail_crit, OPT_INT)
 OPTION(mon_data_avail_warn, OPT_INT)
 OPTION(mon_data_size_warn, OPT_U64) // issue a warning when the monitor's data store goes over 15GB (in bytes)
@@ -345,8 +339,6 @@ OPTION(mon_client_ping_timeout, OPT_DOUBLE)   // fail if we don't hear back
 OPTION(mon_client_hunt_interval_backoff, OPT_DOUBLE) // each time we reconnect to a monitor, double our timeout
 OPTION(mon_client_hunt_interval_max_multiple, OPT_DOUBLE) // up to a max of 10*default (30 seconds)
 OPTION(mon_client_max_log_entries_per_message, OPT_INT)
-OPTION(mon_pool_quota_warn_threshold, OPT_INT) // percent of quota at which to issue warnings
-OPTION(mon_pool_quota_crit_threshold, OPT_INT) // percent of quota at which to issue errors
 OPTION(client_cache_size, OPT_INT)
 OPTION(client_cache_mid, OPT_FLOAT)
 OPTION(client_use_random_mds, OPT_BOOL)

diff --git a/src/common/options.cc b/src/common/options.cc
index 111ab63502ac7bdf14c959a9ffcb1736cefced2f..89005ae358903e3ac2513f9d734dfeecdeb4991c 100644 (file)
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -1008,9 +1008,11 @@ std::vector<Option> get_global_options() {
     .set_default(false)
     .set_description(""),
 
-    Option("mon_stat_smooth_intervals", Option::TYPE_INT, Option::LEVEL_ADVANCED)
+    Option("mon_stat_smooth_intervals", Option::TYPE_UINT, Option::LEVEL_ADVANCED)
     .set_default(6)
-    .set_description(""),
+    .set_min(1)
+    .add_service("mgr")
+    .set_description("number of PGMaps stats over which we calc the average read/write throughput of the whole cluster"),
 
     Option("mon_election_timeout", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
     .set_default(5)
@@ -1050,7 +1052,9 @@ std::vector<Option> get_global_options() {
 
     Option("mon_pg_stuck_threshold", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(60)
-    .set_description(""),
+    .set_description("number of seconds after which pgs can be considered stuck inactive, unclean, etc")
+    .set_long_description("see doc/control.rst under dump_stuck for more info")
+    .add_service("mgr"),
 
     Option("mon_pg_min_inactive", Option::TYPE_UINT, Option::LEVEL_ADVANCED)
     .set_default(1)
@@ -1066,19 +1070,23 @@ std::vector<Option> get_global_options() {
 
     Option("mon_pg_warn_max_object_skew", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
     .set_default(10.0)
-    .set_description(""),
+    .set_description("max skew few average in objects per pg")
+    .add_service("mgr"),
 
     Option("mon_pg_warn_min_objects", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(10000)
-    .set_description(""),
+    .set_description("do not warn below this object #")
+    .add_service("mgr"),
 
     Option("mon_pg_warn_min_pool_objects", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(1000)
-    .set_description(""),
+    .set_description("do not warn on pools below this object #")
+    .add_service("mgr"),
 
     Option("mon_pg_check_down_all_threshold", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
     .set_default(.5)
-    .set_description(""),
+    .set_description("threshold of down osds after which we check all pgs")
+    .add_service("mgr"),
 
     Option("mon_cache_target_full_warn_ratio", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
     .set_default(.66)
@@ -1225,9 +1233,9 @@ std::vector<Option> get_global_options() {
     .set_default(true)
     .set_description("Warn about the health JSON format change in preluminous JSON fields"),
 
-    Option("mon_health_max_detail", Option::TYPE_INT, Option::LEVEL_ADVANCED)
+    Option("mon_health_max_detail", Option::TYPE_UINT, Option::LEVEL_ADVANCED)
     .set_default(50)
-    .set_description(""),
+    .set_description("max detailed pgs to report in health detail"),
 
     Option("mon_health_log_update_period", Option::TYPE_INT, Option::LEVEL_DEV)
     .set_default(5)
@@ -1516,11 +1524,13 @@ std::vector<Option> get_global_options() {
 
     Option("mon_pool_quota_warn_threshold", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(0)
-    .set_description(""),
+    .set_description("percent of quota at which to issue warnings")
+    .add_service("mgr"),
 
     Option("mon_pool_quota_crit_threshold", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(0)
-    .set_description(""),
+    .set_description("percent of quota at which to issue errors")
+    .add_service("mgr"),
 
     Option("crush_location", Option::TYPE_STR, Option::LEVEL_ADVANCED)
     .set_default("")

diff --git a/src/mon/PGMap.cc b/src/mon/PGMap.cc
index 461f02e1e9d26476e013561e4a99604a2616a344..1fbce57c211c69afdae666713a1f3898abcecbd2 100644 (file)
--- a/src/mon/PGMap.cc
+++ b/src/mon/PGMap.cc
@@ -1092,7 +1092,9 @@ void PGMap::apply_incremental(CephContext *cct, const Incremental& inc)
   stamp_delta += delta_t;
 
   pg_sum_delta.stats.add(d.stats);
-  if (pg_sum_deltas.size() > (unsigned)MAX(1, cct ? cct->_conf->mon_stat_smooth_intervals : 1)) {
+  auto smooth_intervals =
+    cct ? cct->_conf->get_val<uint64_t>("mon_stat_smooth_intervals") : 1;
+  if (pg_sum_deltas.size() > smooth_intervals) {
     pg_sum_delta.stats.sub(pg_sum_deltas.front().first.stats);
     stamp_delta -= pg_sum_deltas.front().second;
     pg_sum_deltas.pop_front();
@@ -1902,7 +1904,7 @@ void PGMap::update_delta(
    * average it out.
    */
   result_pool_delta->stats.add(d.stats);
-  size_t s = MAX(1, cct ? cct->_conf->mon_stat_smooth_intervals : 1);
+  size_t s = cct ? cct->_conf->get_val<uint64_t>("mon_stat_smooth_intervals") : 1;
   if (delta_avg_list->size() > s) {
     result_pool_delta->stats.sub(delta_avg_list->front().first.stats);
     *result_ts_delta -= delta_avg_list->front().second;
@@ -2083,7 +2085,7 @@ void PGMap::get_health_checks(
   health_check_map_t *checks) const
 {
   utime_t now = ceph_clock_now();
-  const unsigned max = cct->_conf->mon_health_max_detail;
+  const auto max = cct->_conf->get_val<uint64_t>("mon_health_max_detail");
   const auto& pools = osdmap.get_pools();
 
   typedef enum pg_consequence_t {
@@ -2189,7 +2191,7 @@ void PGMap::get_health_checks(
     }
   }
 
-  utime_t cutoff = now - utime_t(cct->_conf->mon_pg_stuck_threshold, 0);
+  utime_t cutoff = now - utime_t(cct->_conf->get_val<int64_t>("mon_pg_stuck_threshold"), 0);
   // Loop over all PGs, if there are any possibly-unhealthy states in there
   if (!possible_responses.empty()) {
     for (const auto& i : pg_stat) {
@@ -2457,6 +2459,12 @@ void PGMap::get_health_checks(
   // MANY_OBJECTS_PER_PG
   if (!pg_stat.empty()) {
     list<string> pgp_detail, many_detail;
+    const auto mon_pg_warn_min_objects =
+      cct->_conf->get_val<int64_t>("mon_pg_warn_min_objects");
+    const auto mon_pg_warn_min_pool_objects =
+      cct->_conf->get_val<int64_t>("mon_pg_warn_min_pool_objects");
+    const auto mon_pg_warn_max_object_skew =
+      cct->_conf->get_val<double>("mon_pg_warn_max_object_skew");
     for (auto p = pg_pool_sum.begin();
          p != pg_pool_sum.end();
          ++p) {
@@ -2474,13 +2482,12 @@ void PGMap::get_health_checks(
       }
       int average_objects_per_pg = pg_sum.stats.sum.num_objects / pg_stat.size();
       if (average_objects_per_pg > 0 &&
-          pg_sum.stats.sum.num_objects >= cct->_conf->mon_pg_warn_min_objects &&
-          p->second.stats.sum.num_objects >=
-         cct->_conf->mon_pg_warn_min_pool_objects) {
+          pg_sum.stats.sum.num_objects >= mon_pg_warn_min_objects &&
+          p->second.stats.sum.num_objects >= mon_pg_warn_min_pool_objects) {
        int objects_per_pg = p->second.stats.sum.num_objects / pi->get_pg_num();
        float ratio = (float)objects_per_pg / (float)average_objects_per_pg;
-       if (cct->_conf->mon_pg_warn_max_object_skew > 0 &&
-           ratio > cct->_conf->mon_pg_warn_max_object_skew) {
+       if (mon_pg_warn_max_object_skew > 0 &&
+           ratio > mon_pg_warn_max_object_skew) {
          ostringstream ss;
          ss << "pool " << name << " objects per pg ("
             << objects_per_pg << ") is more than " << ratio
@@ -2508,8 +2515,8 @@ void PGMap::get_health_checks(
   // POOL_FULL
   // POOL_NEAR_FULL
   {
-    float warn_threshold = (float)g_conf->mon_pool_quota_warn_threshold/100;
-    float crit_threshold = (float)g_conf->mon_pool_quota_crit_threshold/100;
+    float warn_threshold = (float)g_conf->get_val<int64_t>("mon_pool_quota_warn_threshold")/100;
+    float crit_threshold = (float)g_conf->get_val<int64_t>("mon_pool_quota_crit_threshold")/100;
     list<string> full_detail, nearfull_detail;
     unsigned full_pools = 0, nearfull_pools = 0;
     for (auto it : pools) {
@@ -3024,7 +3031,7 @@ int process_pg_map_command(
       stuckop_vec.push_back("unclean");
     int64_t threshold;
     cmd_getval(g_ceph_context, cmdmap, "threshold", threshold,
-               int64_t(g_conf->mon_pg_stuck_threshold));
+               g_conf->get_val<int64_t>("mon_pg_stuck_threshold"));
 
     r = pg_map.dump_stuck_pg_stats(ds, f, (int)threshold, stuckop_vec);
     odata->append(ds);
@@ -3329,7 +3336,7 @@ void PGMapUpdater::check_down_pgs(
   // if a large number of osds changed state, just iterate over the whole
   // pg map.
   if (need_check_down_pg_osds.size() > (unsigned)osdmap.get_num_osds() *
-      g_conf->mon_pg_check_down_all_threshold) {
+      g_conf->get_val<double>("mon_pg_check_down_all_threshold")) {
     check_all = true;
   }