From: Remi Gacogne <remi.gacogne@powerdns.com>
Date: Wed, 27 Mar 2019 11:10:37 +0000 (+0100)
Subject: rec: Add a 'rebalanced-queries' metric
X-Git-Tag: rec-4.1.12~1^2~3
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=47365ba09d3fe8fc885bf6b2135992a57a02ac6f;p=pdns

rec: Add a 'rebalanced-queries' metric

(cherry picked from commit 596bf48219cb82379bf872e746dd7f626b582342)
---

diff --git a/pdns/pdns_recursor.cc b/pdns/pdns_recursor.cc
index a9f317208..841084575 100644
--- a/pdns/pdns_recursor.cc
+++ b/pdns/pdns_recursor.cc
@@ -2399,9 +2399,13 @@ static unsigned int selectWorker(unsigned int hash)
 
   unsigned int worker = hash % g_numWorkerThreads;
   /* at least one server has to be below the average load */
-  while(load[worker] > targetLoad) {
-    // cerr<<"worker "<<worker<<" is above the target load, selecting another one"<<endl;
-    worker = (worker + 1) % g_numWorkerThreads;
+  if (load[worker] > targetLoad) {
+    ++g_stats.rebalancedQueries;
+    do {
+      // cerr<<"worker "<<worker<<" is above the target load, selecting another one"<<endl;
+      worker = (worker + 1) % g_numWorkerThreads;
+    }
+    while(load[worker] > targetLoad);
   }
 
   return /* skip distributor */ 1 + worker;
diff --git a/pdns/rec-snmp.cc b/pdns/rec-snmp.cc
index 46697550b..f62d8b636 100644
--- a/pdns/rec-snmp.cc
+++ b/pdns/rec-snmp.cc
@@ -292,6 +292,5 @@ RecursorSNMPAgent::RecursorSNMPAgent(const std::string& name, const std::string&
   registerCounter64Stat("policy-result-nodata", policyResultNodataOID, OID_LENGTH(policyResultNodataOID));
   registerCounter64Stat("policy-result-truncate", policyResultTruncateOID, OID_LENGTH(policyResultTruncateOID));
   registerCounter64Stat("policy-result-custom", policyResultCustomOID, OID_LENGTH(policyResultCustomOID));
-
 #endif /* HAVE_NET_SNMP */
 }
diff --git a/pdns/rec_channel_rec.cc b/pdns/rec_channel_rec.cc
index 0ae865e7f..4d5d625b6 100644
--- a/pdns/rec_channel_rec.cc
+++ b/pdns/rec_channel_rec.cc
@@ -988,6 +988,8 @@ void registerAllStats()
   addGetStat("policy-result-nodata", &g_stats.policyResults[DNSFilterEngine::PolicyKind::NODATA]);
   addGetStat("policy-result-truncate", &g_stats.policyResults[DNSFilterEngine::PolicyKind::Truncate]);
   addGetStat("policy-result-custom", &g_stats.policyResults[DNSFilterEngine::PolicyKind::Custom]);
+
+  addGetStat("rebalanced-queries", &g_stats.rebalancedQueries);
 }
 
 static void doExitGeneric(bool nicely)
diff --git a/pdns/recursordist/docs/metrics.rst b/pdns/recursordist/docs/metrics.rst
index a3519010f..70b69cbcd 100644
--- a/pdns/recursordist/docs/metrics.rst
+++ b/pdns/recursordist/docs/metrics.rst
@@ -370,6 +370,12 @@ questions
 ^^^^^^^^^
 counts all end-user initiated queries with the RD bit   set
 
+rebalanced-queries
+^^^^^^^^^^^^^^^^^^
+.. versionadded:: 4.1.12
+
+number of queries balanced to a different worker thread because the first selected one was above the target load configured with 'distribution-load-factor'
+
 resource-limits
 ^^^^^^^^^^^^^^^
 counts number of queries that could not be   performed because of resource limits
diff --git a/pdns/syncres.hh b/pdns/syncres.hh
index 70ef510bb..75e28b05f 100644
--- a/pdns/syncres.hh
+++ b/pdns/syncres.hh
@@ -916,6 +916,7 @@ struct RecursorStats
   std::atomic<uint64_t> dnssecValidations; // should be the sum of all dnssecResult* stats
   std::map<vState, std::atomic<uint64_t> > dnssecResults;
   std::map<DNSFilterEngine::PolicyKind, std::atomic<uint64_t> > policyResults;
+  std::atomic<uint64_t> rebalancedQueries{0};
 };
 
 //! represents a running TCP/IP client session