[PATCH] WL#3704 mgmapi timeouts: For mgm server, have timeout per operation,...

[PATCH] WL#3704 mgmapi timeouts: For mgm server, have timeout per operation, not for entire connection (i.e. fix the bug)

for TransporterRegistry, keep connection to management server alive.


Index: ndb-work/storage/ndb/include/util/InputStream.hpp
===================================================================


storage/ndb/include/util/InputStream.hpp:
  WL#3704 mgmapi timeouts: For mgm server, have timeout per operation, not for entire connection (i.e. fix the bug)
storage/ndb/include/util/OutputStream.hpp:
  WL#3704 mgmapi timeouts: For mgm server, have timeout per operation, not for entire connection (i.e. fix the bug)
storage/ndb/src/common/transporter/TransporterRegistry.cpp:
  WL#3704 mgmapi timeouts: For mgm server, have timeout per operation, not for entire connection (i.e. fix the bug)
storage/ndb/src/mgmsrv/Services.cpp:
  WL#3704 mgmapi timeouts: For mgm server, have timeout per operation, not for entire connection (i.e. fix the bug)

storage/ndb/include/util/InputStream.hpp

@@ -32,6 +32,7 @@ public:
    * Set the mutex to be UNLOCKED when blocking (e.g. select(2))
    */
   void set_mutex(NdbMutex *m) { m_mutex= m; };
+  virtual void reset_timeout() {};
 protected:
   NdbMutex *m_mutex;
 };
@@ -57,7 +58,7 @@ public:
   virtual ~SocketInputStream() {}
   char* gets(char * buf, int bufLen);
   bool timedout() { return m_timedout; };
-  void reset_timeout() { m_timedout= false; };
+  void reset_timeout() { m_timedout= false; m_timeout_remain= m_timeout_ms;};
 
 };
 

storage/ndb/include/util/OutputStream.hpp

@@ -29,6 +29,7 @@ public:
   virtual int print(const char * fmt, ...) = 0;
   virtual int println(const char * fmt, ...) = 0;
   virtual void flush() {};
+  virtual void reset_timeout() {};
 };
 
 class FileOutputStream : public OutputStream {
@@ -51,7 +52,7 @@ public:
   SocketOutputStream(NDB_SOCKET_TYPE socket, unsigned write_timeout_ms = 1000);
   virtual ~SocketOutputStream() {}
   bool timedout() { return m_timedout; };
-  void reset_timeout() { m_timedout= false; };
+  void reset_timeout() { m_timedout= false; m_timeout_remain= m_timeout_ms;};
 
   int print(const char * fmt, ...);
   int println(const char * fmt, ...);

storage/ndb/src/common/transporter/TransporterRegistry.cpp

@@ -1059,9 +1059,16 @@ TransporterRegistry::update_connections()
 void
 TransporterRegistry::start_clients_thread()
 {
+  int persist_mgm_count= 0;
   DBUG_ENTER("TransporterRegistry::start_clients_thread");
   while (m_run_start_clients_thread) {
     NdbSleep_MilliSleep(100);
+    persist_mgm_count++;
+    if(persist_mgm_count==50)
+    {
+      ndb_mgm_check_connection(m_mgm_handle);
+      persist_mgm_count= 0;
+    }
     for (int i= 0, n= 0; n < nTransporters && m_run_start_clients_thread; i++){
       Transporter * t = theTransporters[i];
       if (!t)
@@ -1119,7 +1126,12 @@ TransporterRegistry::start_clients_thread()
 	      {
 		g_eventLogger.info("Management server closed connection early. "
 			 "It is probably being shut down (or has problems). "
-			 "We will retry the connection.");
+			 "We will retry the connection. %d %s %s line: %d",
+                                   ndb_mgm_get_latest_error(m_mgm_handle),
+                                   ndb_mgm_get_latest_error_desc(m_mgm_handle),
+                                   ndb_mgm_get_latest_error_msg(m_mgm_handle),
+                                   ndb_mgm_get_latest_error_line(m_mgm_handle)
+                                   );
 	      }
 	    }
 	    /** else

storage/ndb/src/mgmsrv/Services.cpp

@@ -344,6 +344,9 @@ MgmApiSession::runSession()
   while(!stop) {
     NdbMutex_Lock(m_mutex);
 
+    m_input->reset_timeout();
+    m_output->reset_timeout();
+
     m_parser->run(ctx, *this);
 
     if(ctx.m_currentToken == 0)