postgres-xc-committers Mailing List for Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  8ce8906c2d45e0aa1164c9beaedb2637853a2e03 (commit)
      from  5800b1b7b84dac3759f25a4a37afcb2ed26a1a63 (commit)


- Log -----------------------------------------------------------------
commit 8ce8906c2d45e0aa1164c9beaedb2637853a2e03
Author: Mason S <masonsharp@mason-sharps-macbook.local>
Date:   Wed Jul 7 15:33:45 2010 +0200

    Fix a crash that may occur within the pooler when a
    data node crashes.
    
    Written by Andrei Matsinchyk

diff --git a/src/backend/pgxc/pool/poolmgr.c b/src/backend/pgxc/pool/poolmgr.c
index 79106b5..6427da3 100644
--- a/src/backend/pgxc/pool/poolmgr.c
+++ b/src/backend/pgxc/pool/poolmgr.c
@@ -4,15 +4,15 @@
  *
  *	  Connection pool manager handles connections to DataNodes
  *
- * The pooler runs as a separate process and is forked off from a 
- * coordinator postmaster. If the coordinator needs a connection from a 
+ * The pooler runs as a separate process and is forked off from a
+ * coordinator postmaster. If the coordinator needs a connection from a
  * data node, it asks for one from the pooler, which maintains separate
  * pools for each data node. A group of connections can be requested in
- * a single request, and the pooler returns a list of file descriptors 
+ * a single request, and the pooler returns a list of file descriptors
  * to use for the connections.
  *
  * Note the current implementation does not yet shrink the pool over time
- * as connections are idle.  Also, it does not queue requests; if a 
+ * as connections are idle.  Also, it does not queue requests; if a
  * connection is unavailable, it will simply fail. This should be implemented
  * one day, although there is a chance for deadlocks. For now, limiting
  * connections should be done between the application and coordinator.
@@ -113,8 +113,8 @@ extern int	pqReadReady(PGconn *conn);
 static volatile sig_atomic_t shutdown_requested = false;
 
 
-/* 
- * Initialize internal structures 
+/*
+ * Initialize internal structures
  */
 int
 PoolManagerInit()
@@ -433,8 +433,8 @@ PoolManagerInit()
 }
 
 
-/* 
- * Destroy internal structures 
+/*
+ * Destroy internal structures
  */
 int
 PoolManagerDestroy(void)
@@ -575,8 +575,8 @@ PoolManagerConnect(PoolHandle *handle, const char *database)
 }
 
 
-/* 
- * Init PoolAgent 
+/*
+ * Init PoolAgent
 */
 static void
 agent_init(PoolAgent *agent, const char *database, List *nodes)
@@ -598,8 +598,8 @@ agent_init(PoolAgent *agent, const char *database, List *nodes)
 }
 
 
-/* 
- * Destroy PoolAgent 
+/*
+ * Destroy PoolAgent
  */
 static void
 agent_destroy(PoolAgent *agent)
@@ -636,8 +636,8 @@ agent_destroy(PoolAgent *agent)
 }
 
 
-/* 
- * Release handle to pool manager 
+/*
+ * Release handle to pool manager
  */
 void
 PoolManagerDisconnect(PoolHandle *handle)
@@ -653,8 +653,8 @@ PoolManagerDisconnect(PoolHandle *handle)
 }
 
 
-/* 
- * Get pooled connections 
+/*
+ * Get pooled connections
  */
 int *
 PoolManagerGetConnections(List *nodelist)
@@ -759,7 +759,7 @@ agent_handle_input(PoolAgent * agent, StringInfo s)
 }
 
 
-/* 
+/*
  * acquire connection
  */
 static int *
@@ -827,8 +827,8 @@ agent_acquire_connections(PoolAgent *agent, List *nodelist)
 }
 
 
-/* 
- * Retun connections back to the pool 
+/*
+ * Retun connections back to the pool
  */
 void
 PoolManagerReleaseConnections(void)
@@ -972,8 +972,8 @@ destroy_database_pool(const char *database)
 }
 
 
-/* 
- * Insert new database pool to the list 
+/*
+ * Insert new database pool to the list
  */
 static void
 insert_database_pool(DatabasePool *databasePool)
@@ -991,8 +991,8 @@ insert_database_pool(DatabasePool *databasePool)
 }
 
 
-/* 
- * Find pool for specified database in the list 
+/*
+ * Find pool for specified database in the list
  */
 static DatabasePool
 *
@@ -1015,8 +1015,8 @@ find_database_pool(const char *database)
 }
 
 
-/* 
- * Remove pool for specified database from the list 
+/*
+ * Remove pool for specified database from the list
  */
 static DatabasePool *
 remove_database_pool(const char *database)
@@ -1075,41 +1075,40 @@ acquire_connection(DatabasePool *dbPool, int node)
 	}
 
 	/* Check available connections */
-	if (nodePool && nodePool->freeSize > 0)
+	while (nodePool && nodePool->freeSize > 0)
 	{
 		int			poll_result;
 
-		while (nodePool->freeSize > 0)
-		{
-			slot = nodePool->slot[--(nodePool->freeSize)];
+		slot = nodePool->slot[--(nodePool->freeSize)];
 
 	retry:
-			/* Make sure connection is ok */
-			poll_result = pqReadReady(slot->conn);
-
-			if (poll_result == 0)
-				break; 		/* ok, no data */
-			else if (poll_result < 0)
-			{
-				if (errno == EAGAIN || errno == EINTR)
-					goto retry;
+		/* Make sure connection is ok */
+		poll_result = pqReadReady(slot->conn);
 
-				elog(WARNING, "Error in checking connection, errno = %d", errno);
-			}
-			else
-				elog(WARNING, "Unexpected data on connection, cleaning.");
+		if (poll_result == 0)
+			break; 		/* ok, no data */
+		else if (poll_result < 0)
+		{
+			if (errno == EAGAIN || errno == EINTR)
+				goto retry;
 
-			destroy_slot(slot);
-			/* Decrement current max pool size */
-			(nodePool->size)--;
-			/* Ensure we are not below minimum size */
-			grow_pool(dbPool, node - 1);
+			elog(WARNING, "Error in checking connection, errno = %d", errno);
 		}
+		else
+			elog(WARNING, "Unexpected data on connection, cleaning.");
+
+		destroy_slot(slot);
+		slot = NULL;
+
+		/* Decrement current max pool size */
+		(nodePool->size)--;
+		/* Ensure we are not below minimum size */
+		grow_pool(dbPool, node - 1);
 	}
-	else
-		ereport(LOG,
-				(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
-				 errmsg("connection pool is empty")));
+
+	if (slot == NULL)
+		elog(WARNING, "can not connect to data node %d", node);
+
 	return slot;
 }
 

-----------------------------------------------------------------------

Summary of changes:
 src/backend/pgxc/pool/poolmgr.c |  101 +++++++++++++++++++--------------------
 1 files changed, 50 insertions(+), 51 deletions(-)


hooks/post-receive
-- 
Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  5800b1b7b84dac3759f25a4a37afcb2ed26a1a63 (commit)
       via  73249fbb42f4c05de85181428a7ae143c0c8d254 (commit)
      from  47f4b06f6e25426bb775d7a7372309b68c7e1f47 (commit)


- Log -----------------------------------------------------------------
commit 5800b1b7b84dac3759f25a4a37afcb2ed26a1a63
Author: Mason S <masonsharp@mason-sharps-macbook.local>
Date:   Wed Jul 7 15:31:15 2010 +0200

    In Postgres-XC, the error stack may overflow because
    AbortTransaction may be called multiple times, each
    time calling DataNodeRollback, which may fail again
    if a data node is down.
    
    Instead, if we are already in an abort state, we do not
    bother repeating abort actions.

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 757f99d..491d0d5 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -2697,6 +2697,20 @@ AbortCurrentTransaction(void)
 	}
 }
 
+#ifdef PGXC
+/*
+ *	AbortCurrentTransactionOnce
+ *
+ * Abort transaction, but only if we have not already.
+ */
+void
+AbortCurrentTransactionOnce(void)
+{
+	if (CurrentTransactionState->state != TRANS_ABORT)
+		AbortCurrentTransaction();
+}
+#endif
+
 /*
  *	PreventTransactionChain
  *
diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c
index 4cb0b27..553a682 100644
--- a/src/backend/tcop/postgres.c
+++ b/src/backend/tcop/postgres.c
@@ -3768,7 +3768,15 @@ PostgresMain(int argc, char *argv[], const char *username)
 		/*
 		 * Abort the current transaction in order to recover.
 		 */
+#ifdef PGXC
+		/* 
+		 * Temporarily do not abort if we are already in an abort state.
+		 * This change tries to handle the case where the error data stack fills up.
+		*/
+		AbortCurrentTransactionOnce();
+#else
 		AbortCurrentTransaction();
+#endif
 
 		/*
 		 * Now return to normal top-level context and clear ErrorContext for
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index fe69611..5bd157b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -163,6 +163,9 @@ extern void CommandCounterIncrement(void);
 extern void ForceSyncCommit(void);
 extern void StartTransactionCommand(void);
 extern void CommitTransactionCommand(void);
+#ifdef PGXC
+extern void AbortCurrentTransactionOnce(void);
+#endif
 extern void AbortCurrentTransaction(void);
 extern void BeginTransactionBlock(void);
 extern bool EndTransactionBlock(void);

commit 73249fbb42f4c05de85181428a7ae143c0c8d254
Author: Mason S <masonsharp@mason-sharps-macbook.local>
Date:   Wed Jul 7 15:30:16 2010 +0200

    Changed some error messages so that they will not be duplicates
    to better pinpoint some issues.

diff --git a/src/backend/pgxc/pool/execRemote.c b/src/backend/pgxc/pool/execRemote.c
index 1ee1d59..c6f9042 100644
--- a/src/backend/pgxc/pool/execRemote.c
+++ b/src/backend/pgxc/pool/execRemote.c
@@ -475,7 +475,7 @@ HandleCopyOutComplete(RemoteQueryState *combiner)
 		/* Inconsistent responses */
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'c' message, current request type %d", combiner->request_type)));
 	/* Just do nothing, close message is managed by the coordinator */
 	combiner->copy_out_count++;
 }
@@ -559,7 +559,7 @@ HandleRowDescription(RemoteQueryState *combiner, char *msg_body, size_t len)
 		/* Inconsistent responses */
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'T' message, current request type %d", combiner->request_type)));
 	}
 	/* Increment counter and check if it was first */
 	if (combiner->description_count++ == 0)
@@ -583,7 +583,7 @@ HandleParameterStatus(RemoteQueryState *combiner, char *msg_body, size_t len)
 		/* Inconsistent responses */
 		ereport(ERROR,
 			(errcode(ERRCODE_DATA_CORRUPTED),
-			 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'S' message, current request type %d", combiner->request_type)));
 	}
 	/* Proxy last */
 	if (++combiner->description_count == combiner->node_count)
@@ -605,7 +605,7 @@ HandleCopyIn(RemoteQueryState *combiner)
 		/* Inconsistent responses */
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'G' message, current request type %d", combiner->request_type)));
 	}
 	/*
 	 * The normal PG code will output an G message when it runs in the
@@ -627,7 +627,7 @@ HandleCopyOut(RemoteQueryState *combiner)
 		/* Inconsistent responses */
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'H' message, current request type %d", combiner->request_type)));
 	}
 	/*
 	 * The normal PG code will output an H message when it runs in the
@@ -649,7 +649,7 @@ HandleCopyDataRow(RemoteQueryState *combiner, char *msg_body, size_t len)
 	if (combiner->request_type != REQUEST_TYPE_COPY_OUT)
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'd' message, current request type %d", combiner->request_type)));
 
 	/* If there is a copy file, data has to be sent to the local file */
 	if (combiner->copy_file)
@@ -675,7 +675,7 @@ HandleDataRow(RemoteQueryState *combiner, char *msg_body, size_t len)
 		/* Inconsistent responses */
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes for 'D' message, current request type %d", combiner->request_type)));
 	}
 
 	/*
@@ -943,7 +943,8 @@ data_node_receive_responses(const int conn_count, DataNodeHandle ** connections,
 		data_node_receive(count, to_receive, timeout);
 		while (i < count)
 		{
-			switch (handle_response(to_receive[i], combiner))
+			int result =  handle_response(to_receive[i], combiner);
+			switch (result)
 			{
 				case RESPONSE_EOF: /* have something to read, keep receiving */
 					i++;
@@ -960,7 +961,7 @@ data_node_receive_responses(const int conn_count, DataNodeHandle ** connections,
 					/* Inconsistent responses */
 					ereport(ERROR,
 							(errcode(ERRCODE_INTERNAL_ERROR),
-							 errmsg("Unexpected response from the data nodes")));
+							 errmsg("Unexpected response from the data nodes, result = %d, request type %d", result, combiner->request_type)));
 			}
 		}
 	}
@@ -1679,7 +1680,7 @@ DataNodeCopyOut(Exec_Nodes *exec_nodes, DataNodeHandle** copy_connections, FILE*
 		pfree(copy_connections);
 		ereport(ERROR,
 				(errcode(ERRCODE_DATA_CORRUPTED),
-				 errmsg("Unexpected response from the data nodes")));
+				 errmsg("Unexpected response from the data nodes when combining, request type %d", combiner->request_type)));
 	}
 
 	return processed;

-----------------------------------------------------------------------

Summary of changes:
 src/backend/access/transam/xact.c  |   14 ++++++++++++++
 src/backend/pgxc/pool/execRemote.c |   21 +++++++++++----------
 src/backend/tcop/postgres.c        |    8 ++++++++
 src/include/access/xact.h          |    3 +++
 4 files changed, 36 insertions(+), 10 deletions(-)


hooks/post-receive
-- 
Postgres-XC