pdns/dnsdist-carbon.cc

   1 /*
   2  * This file is part of PowerDNS or dnsdist.
   3  * Copyright -- PowerDNS.COM B.V. and its contributors
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of version 2 of the GNU General Public License as
   7  * published by the Free Software Foundation.
   8  *
   9  * In addition, for the avoidance of any doubt, permission is granted to
  10  * link this program with OpenSSL and to (re)distribute the binaries
  11  * produced as the result of such linking.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * along with this program; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include "config.h"
  24 #endif
  25 #include "iputils.hh"
  26 #include "dolog.hh"
  27 #include "sstuff.hh"
  28
  29 #include "namespaces.hh"
  30 #include "dnsdist.hh"
  31 #include "threadname.hh"
  32
  33 GlobalStateHolder<vector<CarbonConfig> > g_carbon;
  34 static time_t s_start=time(0);
  35 uint64_t uptimeOfProcess(const std::string& str)
  36 {
  37   return time(0) - s_start;
  38 }
  39
  40 void carbonDumpThread()
  41 try
  42 {
  43   setThreadName("dnsdist/carbon");
  44   auto localCarbon = g_carbon.getLocal();
  45   for(int numloops=0;;++numloops) {
  46     if(localCarbon->empty()) {
  47       sleep(1);
  48       continue;
  49     }
  50     /* this is wrong, we use the interval of the first server
  51        for every single one of them */
  52     if(numloops) {
  53       const unsigned int interval = localCarbon->at(0).interval;
  54       sleep(interval);
  55     }
  56
  57     for (const auto& conf : *localCarbon) {
  58       const auto& server = conf.server;
  59       const std::string& namespace_name = conf.namespace_name;
  60       std::string hostname = conf.ourname;
  61       if(hostname.empty()) {
  62         char tmp[80];
  63         memset(tmp, 0, sizeof(tmp));
  64         gethostname(tmp, sizeof(tmp));
  65         char *p = strchr(tmp, '.');
  66         if(p) *p=0;
  67         hostname=tmp;
  68         boost::replace_all(hostname, ".", "_");
  69       }
  70       const std::string& instance_name = conf.instance_name;
  71
  72       try {
  73         Socket s(server.sin4.sin_family, SOCK_STREAM);
  74         s.setNonBlocking();
  75         s.connect(server);  // we do the connect so the attempt happens while we gather stats
  76         ostringstream str;
  77         time_t now=time(0);
  78         for(const auto& e : g_stats.entries) {
  79           str<<namespace_name<<"."<<hostname<<"."<<instance_name<<"."<<e.first<<' ';
  80           if(const auto& val = boost::get<DNSDistStats::stat_t*>(&e.second))
  81             str<<(*val)->load();
  82           else if (const auto& dval = boost::get<double*>(&e.second))
  83             str<<**dval;
  84           else
  85             str<<(*boost::get<DNSDistStats::statfunction_t>(&e.second))(e.first);
  86           str<<' '<<now<<"\r\n";
  87         }
  88         auto states = g_dstates.getLocal();
  89         for(const auto& state : *states) {
  90           string serverName = state->name.empty() ? (state->remote.toString() + ":" + std::to_string(state->remote.getPort())) : state->getName();
  91           boost::replace_all(serverName, ".", "_");
  92           const string base = namespace_name + "." + hostname + "." + instance_name + ".servers." + serverName + ".";
  93           str<<base<<"queries" << ' ' << state->queries.load() << " " << now << "\r\n";
  94           str<<base<<"drops" << ' ' << state->reuseds.load() << " " << now << "\r\n";
  95           str<<base<<"latency" << ' ' << (state->availability != DownstreamState::Availability::Down ? state->latencyUsec/1000.0 : 0) << " " << now << "\r\n";
  96           str<<base<<"senderrors" << ' ' << state->sendErrors.load() << " " << now << "\r\n";
  97           str<<base<<"outstanding" << ' ' << state->outstanding.load() << " " << now << "\r\n";
  98           str<<base<<"tcpdiedsendingquery" << ' '<< state->tcpDiedSendingQuery.load() << " " << now << "\r\n";
  99           str<<base<<"tcpdiedreaddingresponse" << ' '<< state->tcpDiedReadingResponse.load() << " " << now << "\r\n";
 100           str<<base<<"tcpgaveup" << ' '<< state->tcpGaveUp.load() << " " << now << "\r\n";
 101           str<<base<<"tcpreadimeouts" << ' '<< state->tcpReadTimeouts.load() << " " << now << "\r\n";
 102           str<<base<<"tcpwritetimeouts" << ' '<< state->tcpWriteTimeouts.load() << " " << now << "\r\n";
 103           str<<base<<"tcpcurrentconnections" << ' '<< state->tcpCurrentConnections.load() << " " << now << "\r\n";
 104           str<<base<<"tcpavgqueriesperconnection" << ' '<< state->tcpAvgQueriesPerConnection.load() << " " << now << "\r\n";
 105           str<<base<<"tcpavgconnectionduration" << ' '<< state->tcpAvgConnectionDuration.load() << " " << now << "\r\n";
 106         }
 107         for(const auto& front : g_frontends) {
 108           if (front->udpFD == -1 && front->tcpFD == -1)
 109             continue;
 110
 111           string frontName = front->local.toString() + ":" + std::to_string(front->local.getPort()) +  (front->udpFD >= 0 ? "_udp" : "_tcp");
 112           boost::replace_all(frontName, ".", "_");
 113           const string base = namespace_name + "." + hostname + "." + instance_name + ".frontends." + frontName + ".";
 114           str<<base<<"queries" << ' ' << front->queries.load() << " " << now << "\r\n";
 115           str<<base<<"tcpdiedreadingquery" << ' '<< front->tcpDiedReadingQuery.load() << " " << now << "\r\n";
 116           str<<base<<"tcpdiedsendingresponse" << ' '<< front->tcpDiedSendingResponse.load() << " " << now << "\r\n";
 117           str<<base<<"tcpgaveup" << ' '<< front->tcpGaveUp.load() << " " << now << "\r\n";
 118           str<<base<<"tcpclientimeouts" << ' '<< front->tcpClientTimeouts.load() << " " << now << "\r\n";
 119           str<<base<<"tcpdownstreamtimeouts" << ' '<< front->tcpDownstreamTimeouts.load() << " " << now << "\r\n";
 120           str<<base<<"tcpcurrentconnections" << ' '<< front->tcpCurrentConnections.load() << " " << now << "\r\n";
 121           str<<base<<"tcpavgqueriesperconnection" << ' '<< front->tcpAvgQueriesPerConnection.load() << " " << now << "\r\n";
 122           str<<base<<"tcpavgconnectionduration" << ' '<< front->tcpAvgConnectionDuration.load() << " " << now << "\r\n";
 123         }
 124         auto localPools = g_pools.getLocal();
 125         for (const auto& entry : *localPools) {
 126           string poolName = entry.first;
 127           boost::replace_all(poolName, ".", "_");
 128           if (poolName.empty()) {
 129             poolName = "_default_";
 130           }
 131           const string base = namespace_name + "." + hostname + "." + instance_name + ".pools." + poolName + ".";
 132           const std::shared_ptr<ServerPool> pool = entry.second;
 133           str<<base<<"servers" << " " << pool->countServers(false) << " " << now << "\r\n";
 134           str<<base<<"servers-up" << " " << pool->countServers(true) << " " << now << "\r\n";
 135           if (pool->packetCache != nullptr) {
 136             const auto& cache = pool->packetCache;
 137             str<<base<<"cache-size" << " " << cache->getMaxEntries() << " " << now << "\r\n";
 138             str<<base<<"cache-entries" << " " << cache->getEntriesCount() << " " << now << "\r\n";
 139             str<<base<<"cache-hits" << " " << cache->getHits() << " " << now << "\r\n";
 140             str<<base<<"cache-misses" << " " << cache->getMisses() << " " << now << "\r\n";
 141             str<<base<<"cache-deferred-inserts" << " " << cache->getDeferredInserts() << " " << now << "\r\n";
 142             str<<base<<"cache-deferred-lookups" << " " << cache->getDeferredLookups() << " " << now << "\r\n";
 143             str<<base<<"cache-lookup-collisions" << " " << cache->getLookupCollisions() << " " << now << "\r\n";
 144             str<<base<<"cache-insert-collisions" << " " << cache->getInsertCollisions() << " " << now << "\r\n";
 145             str<<base<<"cache-ttl-too-shorts" << " " << cache->getTTLTooShorts() << " " << now << "\r\n";
 146           }
 147         }
 148
 149 #ifdef HAVE_DNS_OVER_HTTPS
 150         {
 151           const string base = "dnsdist." + hostname + ".main.doh.";
 152           for(const auto& doh : g_dohlocals) {
 153             string name = doh->d_local.toStringWithPort();
 154             boost::replace_all(name, ".", "_");
 155             boost::replace_all(name, ":", "_");
 156             boost::replace_all(name, "[", "_");
 157             boost::replace_all(name, "]", "_");
 158
 159             vector<pair<const char*, const std::atomic<uint64_t>&>> v{
 160               {"http-connects", doh->d_httpconnects},
 161               {"http1-queries", doh->d_http1queries},
 162               {"http2-queries", doh->d_http2queries},
 163               {"tls10-queries", doh->d_tls10queries},
 164               {"tls11-queries", doh->d_tls11queries},
 165               {"tls12-queries", doh->d_tls12queries},
 166               {"tls13-queries", doh->d_tls13queries},
 167               {"tls-unknown-queries", doh->d_tlsUnknownqueries},
 168               {"get-queries", doh->d_getqueries},
 169               {"post-queries", doh->d_postqueries},
 170               {"bad-requests", doh->d_badrequests},
 171               {"error-responses", doh->d_errorresponses},
 172               {"valid-responses", doh->d_validresponses}
 173             };
 174
 175             for(const auto& item : v) {
 176               str<<base<<name<<"."<<item.first << " " << item.second << " " << now <<"\r\n";
 177             }
 178           }
 179         }
 180 #endif /* HAVE_DNS_OVER_HTTPS */
 181
 182         {
 183           WriteLock wl(&g_qcount.queryLock);
 184           std::string qname;
 185           for(auto &record: g_qcount.records) {
 186             qname = record.first;
 187             boost::replace_all(qname, ".", "_");
 188             str<<"dnsdist.querycount." << qname << ".queries " << record.second << " " << now << "\r\n";
 189           }
 190           g_qcount.records.clear();
 191         }
 192
 193         const string msg = str.str();
 194
 195         int ret = waitForRWData(s.getHandle(), false, 1 , 0);
 196         if(ret <= 0 ) {
 197           vinfolog("Unable to write data to carbon server on %s: %s", server.toStringWithPort(), (ret<0 ? strerror(errno) : "Timeout"));
 198           continue;
 199         }
 200         s.setBlocking();
 201         writen2(s.getHandle(), msg.c_str(), msg.size());
 202       }
 203       catch(std::exception& e) {
 204         warnlog("Problem sending carbon data: %s", e.what());
 205       }
 206     }
 207   }
 208 }
 209 catch(std::exception& e)
 210 {
 211   errlog("Carbon thread died: %s", e.what());
 212 }
 213 catch(PDNSException& e)
 214 {
 215   errlog("Carbon thread died, PDNSException: %s", e.reason);
 216 }
 217 catch(...)
 218 {
 219   errlog("Carbon thread died");
 220 }