port 3.4 changes that give libcurl time to retry a failed DNS query. Also add changes to vocbase.cpp that were missed in previous PR. (#7132)

CHANGELOG

@@ -1,6 +1,9 @@
 devel
 -----
 
+* force connection timeout to be 7 seconds to allow libcurl time to retry lost DNS
+  queries.
+
 * fixes a routing issue within the web ui after the use of views
 
 * fixes some graph data parsing issues in the ui, e.g. cleaning up duplicate

arangod/VocBase/vocbase.cpp

@@ -872,7 +872,10 @@ void TRI_vocbase_t::shutdown() {
 
   // starts unloading of collections
   for (auto& collection : collections) {
-    collection->close(); // required to release indexes
+    {
+      WRITE_LOCKER_EVENTUAL(locker, collection->lock());
+      collection->close(); // required to release indexes
+    }
     unloadCollection(collection.get(), true);
   }
 
@@ -1828,6 +1831,7 @@ TRI_vocbase_t::~TRI_vocbase_t() {
 
   // do a final cleanup of collections
   for (auto& it : _collections) {
+    WRITE_LOCKER_EVENTUAL(locker, it->lock());
     it->close(); // required to release indexes
   }
 

lib/SimpleHttpClient/Communicator.cpp

@@ -382,8 +382,11 @@ void Communicator::createRequestInProgress(NewRequest&& newRequest) {
   // in doubt change the timeout to _MS below and hardcode it to 999 and see if
   // the requests immediately fail
   // if not this hack can go away
-  if (connectTimeout <= 0) {
+  if (connectTimeout <= 7) {
-    connectTimeout = 5;
+    // matthewv: previously arangod default was 1.  libcurl flushes its DNS cache
+    //  every 60 seconds.  Tests showed DNS packets lost under high load.  libcurl
+    //  retries DNS after 5 seconds.  7 seconds allows for one retry plus a little padding.
+    connectTimeout = 7;
   }
 
   curl_easy_setopt(
@@ -485,6 +488,14 @@ void Communicator::handleResult(CURL* handle, CURLcode rc) {
           << ::buildPrefix(rip->_ticketId) << "curl error details: " << rip->_errorBuffer;
       }
 
+      double namelookup;
+      curl_easy_getinfo(handle, CURLINFO_NAMELOOKUP_TIME, &namelookup);
+
+      if (5.0 <= namelookup) {
+        LOG_TOPIC(WARN, arangodb::Logger::FIXME) << "libcurl DNS lookup took "
+                                                 << namelookup << " seconds.  Consider using static IP addresses.";
+      } // if
+
       switch (rc) {
         case CURLE_OK: {
           long httpStatusCode = 200;