Fix recovery completed message handling

knizhnik · kelvich · commit 025f3fdfa9f1 · 2017-01-25T12:38:48.000+03:00
diff --git a/contrib/mmts/multimaster.c b/contrib/mmts/multimaster.c
@@ -3298,6 +3298,7 @@ void MtmRecoverNode(int nodeId)
 	{
 		Assert(BIT_CHECK(Mtm->disabledNodeMask, nodeId-1));
 		BIT_CLEAR(Mtm->stoppedNodeMask, nodeId-1);
+		BIT_CLEAR(Mtm->stalledNodeMask, nodeId-1);
 	}
 	MtmUnlock();
 		
@@ -3556,8 +3557,8 @@ MtmReplicationRowFilterHook(struct PGLogicalRowFilterArgs* args)
  * Filter received transactions at destination side.
  * This function is executed by receiver, 
  * so there are no race conditions and it is possible to update nodes[i].restartLSN without lock. 
- * It is more efficient to filter records at senders size (done by MtmReplicationTxnFilterHook) to avoid sending useless data through network. But asynchronous nature of 
- * logical replications makes it not possible to guarantee (at least I failed to do it)
+ * It is more efficient to filter records at senders size (done by MtmReplicationTxnFilterHook) to avoid sending useless data through network. 
+ * But asynchronous nature of logical replications makes it not possible to guarantee (at least I failed to do it)
  * that replica do not receive deteriorated data.
  */
 bool MtmFilterTransaction(char* record, int size)
diff --git a/contrib/mmts/pglogical_apply.c b/contrib/mmts/pglogical_apply.c
@@ -1119,7 +1119,7 @@ void MtmExecutor(void* work, size_t size)
 			case 'Z':
 			{
 				MtmRecoveryCompleted();
-				continue;
+				break;
 			}
             default:
                 elog(ERROR, "unknown action of type %c", action);
diff --git a/contrib/mmts/pglogical_proto.c b/contrib/mmts/pglogical_proto.c
@@ -182,7 +182,7 @@ pglogical_write_commit(StringInfo out, PGLogicalOutputData *data,
 {
     uint8 event = 0;
 	
-	MTM_LOG1("%d: pglogical_write_commit XID=%ld node=%d restart_decoding_lsn=%lx first_lsn=%lx end_lsn=%lx confirmed_flush=%lx", 
+	MTM_LOG2("%d: pglogical_write_commit XID=%ld node=%d restart_decoding_lsn=%lx first_lsn=%lx end_lsn=%lx confirmed_flush=%lx", 
 			 MyProcPid, (long)txn->xid, MtmReplicationNodeId, txn->restart_decoding_lsn, txn->first_lsn, txn->end_lsn, MyReplicationSlot->data.confirmed_flush);
 
 
diff --git a/contrib/mmts/tests2/test_recovery.py b/contrib/mmts/tests2/test_recovery.py
@@ -87,7 +87,7 @@ def tearDownClass(self):
         print('tearDown')
         self.client.stop()
         # XXX: check nodes data identity here
-        subprocess.check_call(['docker-compose','down'])
+#        subprocess.check_call(['docker-compose','down'])
 
     def setUp(self):
         warnings.simplefilter("ignore", ResourceWarning)

Original file line number	Diff line number	Diff line change
`@@ -1119,7 +1119,7 @@ void MtmExecutor(void* work, size_t size)`
`1119`	`1119`	`case 'Z':`
`1120`	`1120`	`{`
`1121`	`1121`	`MtmRecoveryCompleted();`
`1122`		`- continue;`
	`1122`	`+ break;`
`1123`	`1123`	`}`
`1124`	`1124`	`default:`
`1125`	`1125`	`elog(ERROR, "unknown action of type %c", action);`
Original file line number	Diff line number	Diff line change
`@@ -182,7 +182,7 @@ pglogical_write_commit(StringInfo out, PGLogicalOutputData *data,`
`182`	`182`	`{`
`183`	`183`	`uint8 event = 0;`
`184`	`184`
`185`		`- MTM_LOG1("%d: pglogical_write_commit XID=%ld node=%d restart_decoding_lsn=%lx first_lsn=%lx end_lsn=%lx confirmed_flush=%lx",`
	`185`	`+ MTM_LOG2("%d: pglogical_write_commit XID=%ld node=%d restart_decoding_lsn=%lx first_lsn=%lx end_lsn=%lx confirmed_flush=%lx",`
`186`	`186`	`MyProcPid, (long)txn->xid, MtmReplicationNodeId, txn->restart_decoding_lsn, txn->first_lsn, txn->end_lsn, MyReplicationSlot->data.confirmed_flush);`
`187`	`187`
`188`	`188`