dnsdist: Reuse the most recently used TCP conn first, check liveness

diff --git a/pdns/dnsdist-tcp.cc b/pdns/dnsdist-tcp.cc
index ae31db31c0b4c15811d18c7372a08044d1c32e93..bc13210d767ebe881b311788d4c75c4f63796ae4 100644 (file)
--- a/pdns/dnsdist-tcp.cc
+++ b/pdns/dnsdist-tcp.cc
@@ -76,16 +76,31 @@ public:
   static std::shared_ptr<TCPConnectionToBackend> getConnectionToDownstream(std::unique_ptr<FDMultiplexer>& mplexer, std::shared_ptr<DownstreamState>& ds, const struct timeval& now)
   {
     std::shared_ptr<TCPConnectionToBackend> result;
+    struct timeval freshCutOff = now;
+    freshCutOff.tv_sec -= 1;
 
     const auto& it = t_downstreamConnections.find(ds);
     if (it != t_downstreamConnections.end()) {
       auto& list = it->second;
-      if (!list.empty()) {
-        result = std::move(list.front());
-        list.pop_front();
+      while (!list.empty()) {
+        result = std::move(list.back());
+        list.pop_back();
+
         result->setReused();
-        ++ds->tcpReusedConnections;
-        return result;
+        /* for connections that have not been used very recently,
+           check whether they have been closed in the meantime */
+        if (freshCutOff < result->getLastDataReceivedTime()) {
+          /* used recently enough, skip the check */
+          ++ds->tcpReusedConnections;
+          return result;
+        }
+
+        if (isTCPSocketUsable(result->getHandle())) {
+          ++ds->tcpReusedConnections;
+          return result;
+        }
+
+        /* otherwise let's try the next one, if any */
       }
     }
 
@@ -114,11 +129,25 @@ public:
     }
   }
 
-  static void cleanupClosedTCPConnections()
+  static void cleanupClosedTCPConnections(struct timeval now)
   {
+    struct timeval freshCutOff = now;
+    freshCutOff.tv_sec -= 1;
+
     for (auto dsIt = t_downstreamConnections.begin(); dsIt != t_downstreamConnections.end(); ) {
       for (auto connIt = dsIt->second.begin(); connIt != dsIt->second.end(); ) {
-        if (*connIt && isTCPSocketUsable((*connIt)->getHandle())) {
+        if (!(*connIt)) {
+          ++connIt;
+          continue;
+        }
+
+        /* don't bother checking freshly used connections */
+        if (freshCutOff < (*connIt)->getLastDataReceivedTime()) {
+          ++connIt;
+          continue;
+        }
+
+        if (isTCPSocketUsable((*connIt)->getHandle())) {
           ++connIt;
         }
         else {
@@ -1099,7 +1128,7 @@ static void tcpClientThread(int pipefd)
     data.mplexer->run(&now);
 
     if (g_downstreamTCPCleanupInterval > 0 && (now.tv_sec > (lastTCPCleanup + g_downstreamTCPCleanupInterval))) {
-      DownstreamConnectionsManager::cleanupClosedTCPConnections();
+      DownstreamConnectionsManager::cleanupClosedTCPConnections(now);
       lastTCPCleanup = now.tv_sec;
 
       if (g_tcpStatesDumpRequested > 0) {

diff --git a/pdns/dnsdistdist/dnsdist-tcp-downstream.cc b/pdns/dnsdistdist/dnsdist-tcp-downstream.cc
index 3318a1e23e2dd716f4aa2aadfe01da81416f385d..6d6d8aad1f4502246341e2184a44d4cec18ebd07 100644 (file)
--- a/pdns/dnsdistdist/dnsdist-tcp-downstream.cc
+++ b/pdns/dnsdistdist/dnsdist-tcp-downstream.cc
@@ -123,6 +123,7 @@ void TCPConnectionToBackend::handleIO(std::shared_ptr<TCPConnectionToBackend>& c
           conn->d_responseBuffer.reserve(conn->d_responseSize + /* we will need to prepend the size later */ 2);
           conn->d_responseBuffer.resize(conn->d_responseSize);
           conn->d_currentPos = 0;
+          conn->d_lastDataReceivedTime = now;
         }
         else if (conn->d_state == State::waitingForResponseFromBackend && conn->d_currentPos > 0) {
           conn->d_state = State::readingResponseSizeFromBackend;
@@ -135,6 +136,7 @@ void TCPConnectionToBackend::handleIO(std::shared_ptr<TCPConnectionToBackend>& c
         if (iostate == IOState::Done) {
           DEBUGLOG("got response from backend");
           try {
+            conn->d_lastDataReceivedTime = now;
             iostate = conn->handleResponse(conn, now);
           }
           catch (const std::exception& e) {

diff --git a/pdns/dnsdistdist/dnsdist-tcp-downstream.hh b/pdns/dnsdistdist/dnsdist-tcp-downstream.hh
index d582ef8fb05f926cf09cf89b6f79b698110dbe55..4c5e7da44ca5fe691ca335ba1464be407e65b146 100644 (file)
--- a/pdns/dnsdistdist/dnsdist-tcp-downstream.hh
+++ b/pdns/dnsdistdist/dnsdist-tcp-downstream.hh
@@ -169,6 +169,11 @@ public:
 
   void setProxyProtocolValuesSent(std::unique_ptr<std::vector<ProxyProtocolValue>>&& proxyProtocolValuesSent);
 
+  struct timeval getLastDataReceivedTime() const
+  {
+    return d_lastDataReceivedTime;
+  }
+
   std::string toString() const
   {
     ostringstream o;
@@ -238,6 +243,7 @@ private:
   std::shared_ptr<IncomingTCPConnectionState> d_clientConn;
   TCPQuery d_currentQuery;
   struct timeval d_connectionStartTime;
+  struct timeval d_lastDataReceivedTime;
   size_t d_currentPos{0};
   uint64_t d_queries{0};
   uint64_t d_downstreamFailures{0};