ceph/src/mgr/ClusterState.cc

   1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
   2 // vim: ts=8 sw=2 smarttab
   3 /*
   4  * Ceph - scalable distributed file system
   5  *
   6  * Copyright (C) 2014 John Spray <john.spray@inktank.com>
   7  *
   8  * This is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License version 2.1, as published by the Free Software
  11  * Foundation.  See file COPYING.
  12  */
  13
  14 #include "messages/MMgrDigest.h"
  15 #include "messages/MMonMgrReport.h"
  16 #include "messages/MPGStats.h"
  17
  18 #include "mgr/ClusterState.h"
  19
  20 #define dout_context g_ceph_context
  21 #define dout_subsys ceph_subsys_mgr
  22 #undef dout_prefix
  23 #define dout_prefix *_dout << "mgr " << __func__ << " "
  24
  25 ClusterState::ClusterState(MonClient *monc_, Objecter *objecter_)
  26   : monc(monc_), objecter(objecter_), lock("ClusterState"), pgservice(pg_map)
  27 {}
  28
  29 void ClusterState::set_objecter(Objecter *objecter_)
  30 {
  31   Mutex::Locker l(lock);
  32
  33   objecter = objecter_;
  34 }
  35
  36 void ClusterState::set_fsmap(FSMap const &new_fsmap)
  37 {
  38   Mutex::Locker l(lock);
  39
  40   fsmap = new_fsmap;
  41 }
  42
  43 void ClusterState::load_digest(MMgrDigest *m)
  44 {
  45   health_json = std::move(m->health_json);
  46   mon_status_json = std::move(m->mon_status_json);
  47 }
  48
  49 void ClusterState::ingest_pgstats(MPGStats *stats)
  50 {
  51   Mutex::Locker l(lock);
  52
  53   const int from = stats->get_orig_source().num();
  54   bool is_in = false;
  55   objecter->with_osdmap([&is_in, from](const OSDMap &osd_map){
  56       is_in = osd_map.is_in(from);
  57   });
  58
  59   if (is_in) {
  60     pending_inc.update_stat(from, stats->epoch, std::move(stats->osd_stat));
  61   } else {
  62     pending_inc.update_stat(from, stats->epoch, osd_stat_t());
  63   }
  64
  65   for (auto p : stats->pg_stat) {
  66     pg_t pgid = p.first;
  67     const auto &pg_stats = p.second;
  68
  69     // In case we're hearing about a PG that according to last
  70     // OSDMap update should not exist
  71     if (existing_pools.count(pgid.pool()) == 0) {
  72       dout(15) << " got " << pgid
  73                << " reported at " << pg_stats.reported_epoch << ":"
  74                << pg_stats.reported_seq
  75                << " state " << pg_state_string(pg_stats.state)
  76                << " but pool not in " << existing_pools
  77                << dendl;
  78       continue;
  79     }
  80     // In case we already heard about more recent stats from this PG
  81     // from another OSD
  82     if (pg_map.pg_stat[pgid].get_version_pair() > pg_stats.get_version_pair()) {
  83       dout(15) << " had " << pgid << " from "
  84                << pg_map.pg_stat[pgid].reported_epoch << ":"
  85                << pg_map.pg_stat[pgid].reported_seq << dendl;
  86       continue;
  87     }
  88
  89     pending_inc.pg_stat_updates[pgid] = pg_stats;
  90   }
  91 }
  92
  93 void ClusterState::update_delta_stats()
  94 {
  95   pending_inc.stamp = ceph_clock_now();
  96   pending_inc.version = pg_map.version + 1; // to make apply_incremental happy
  97   dout(10) << " v" << pending_inc.version << dendl;
  98
  99   dout(30) << " pg_map before:\n";
 100   JSONFormatter jf(true);
 101   jf.dump_object("pg_map", pg_map);
 102   jf.flush(*_dout);
 103   *_dout << dendl;
 104   dout(30) << " incremental:\n";
 105   JSONFormatter jf(true);
 106   jf.dump_object("pending_inc", pending_inc);
 107   jf.flush(*_dout);
 108   *_dout << dendl;
 109
 110   pg_map.apply_incremental(g_ceph_context, pending_inc);
 111   pending_inc = PGMap::Incremental();
 112 }
 113
 114 void ClusterState::notify_osdmap(const OSDMap &osd_map)
 115 {
 116   Mutex::Locker l(lock);
 117
 118   pending_inc.stamp = ceph_clock_now();
 119   pending_inc.version = pg_map.version + 1; // to make apply_incremental happy
 120   dout(10) << " v" << pending_inc.version << dendl;
 121
 122   PGMapUpdater::check_osd_map(g_ceph_context, osd_map, pg_map, &pending_inc);
 123
 124   // update our list of pools that exist, so that we can filter pg_map updates
 125   // in synchrony with this OSDMap.
 126   existing_pools.clear();
 127   for (auto& p : osd_map.get_pools()) {
 128     existing_pools.insert(p.first);
 129   }
 130
 131   // brute force this for now (don't bother being clever by only
 132   // checking osds that went up/down)
 133   set<int> need_check_down_pg_osds;
 134   PGMapUpdater::check_down_pgs(osd_map, pg_map, true,
 135                                need_check_down_pg_osds, &pending_inc);
 136
 137   dout(30) << " pg_map before:\n";
 138   JSONFormatter jf(true);
 139   jf.dump_object("pg_map", pg_map);
 140   jf.flush(*_dout);
 141   *_dout << dendl;
 142   dout(30) << " incremental:\n";
 143   JSONFormatter jf(true);
 144   jf.dump_object("pending_inc", pending_inc);
 145   jf.flush(*_dout);
 146   *_dout << dendl;
 147
 148   pg_map.apply_incremental(g_ceph_context, pending_inc);
 149   pending_inc = PGMap::Incremental();
 150   // TODO: Complete the separation of PG state handling so
 151   // that a cut-down set of functionality remains in PGMonitor
 152   // while the full-blown PGMap lives only here.
 153 }