1) Drop colocated shards
2) Swap Metadata
3) Drop dummy shards
users/saawasek/non_blocking_split_integrated
Sameer Awasekar 2022-07-14 16:23:00 +05:30
parent 662a4aaec1
commit 2fa1dac051
10 changed files with 750 additions and 428 deletions

View File

@ -81,6 +81,7 @@ LookupSplitMode(Oid shardTransferModeOid)
{
shardSplitMode = BLOCKING_SPLIT;
}
/* Extend with other modes as we support them */
else if (strncmp(enumLabel, "auto", NAMEDATALEN) == 0 ||
strncmp(enumLabel, "force_logical", NAMEDATALEN) == 0)

View File

@ -94,6 +94,10 @@ static void TryDropSplitShardsOnFailure(List *shardGroupSplitIntervalListList,
static Task * CreateTaskForDDLCommandList(uint64 jobId, List *ddlCommandList,
WorkerNode *workerNode);
static void AddDummyShardEntryInMap(uint32 targetNodeId, ShardInterval *shardInterval);
static void DropDummyShards(void);
void TryDropShard(MultiConnection *connection, ShardInterval *shardInterval);
/* Customize error message strings based on operation type */
static const char *const SplitOperationName[] =
{
@ -106,6 +110,9 @@ static const char *const SplitTargetName[] =
[ISOLATE_TENANT_TO_NEW_SHARD] = "tenant",
};
/* Map containing list of dummy shards created on target nodes */
static HTAB *DummyShardInfoHashMap = NULL;
/* Function definitions */
/*
@ -1039,7 +1046,7 @@ TryDropSplitShardsOnFailure(List *shardGroupSplitIntervalListList,
int connectionFlags = FOR_DDL;
connectionFlags |= OUTSIDE_TRANSACTION;
MultiConnection *connnection = GetNodeUserDatabaseConnection(
MultiConnection *connection = GetNodeUserDatabaseConnection(
connectionFlags,
workerPlacementNode->workerName,
workerPlacementNode->workerPort,
@ -1051,13 +1058,14 @@ TryDropSplitShardsOnFailure(List *shardGroupSplitIntervalListList,
* The shard may or may not exist and the connection could have died.
*/
ExecuteOptionalRemoteCommand(
connnection,
connection,
dropShardQuery->data,
NULL /* pgResult */);
}
}
}
/*
* SplitShard API to split a given shard (or shard group) in blocking fashion
* based on specified split points to a set of destination nodes.
@ -1107,12 +1115,33 @@ NonBlockingShardSplit(SplitOperation splitOperation,
sourceShardToCopyNode,
workersForPlacementList);
/*TODO: Refactor this method. BlockWrites is within this as of now, take it out */
SplitShardReplicationSetup(
sourceColocatedShardIntervalList,
shardGroupSplitIntervalListList,
sourceShardToCopyNode,
workersForPlacementList);
/*
* Drop old shards and delete related metadata. Have to do that before
* creating the new shard metadata, because there's cross-checks
* preventing inconsistent metadata (like overlapping shards).
*/
DropShardList(sourceColocatedShardIntervalList);
/* Insert new shard and placement metdata */
InsertSplitChildrenShardMetadata(shardGroupSplitIntervalListList,
workersForPlacementList);
/*
* Create foreign keys if exists after the metadata changes happening in
* DropShardList() and InsertSplitChildrenShardMetadata() because the foreign
* key creation depends on the new metadata.
*/
CreateForeignKeyConstraints(shardGroupSplitIntervalListList,
workersForPlacementList);
DropDummyShards();
}
PG_CATCH();
{
@ -1120,11 +1149,14 @@ NonBlockingShardSplit(SplitOperation splitOperation,
TryDropSplitShardsOnFailure(shardGroupSplitIntervalListList,
workersForPlacementList);
DropDummyShards();
PG_RE_THROW();
}
PG_END_TRY();
}
/* Create ShardGroup split children on a list of corresponding workers. */
static void
CreateSplitShardsForShardGroupTwo(WorkerNode *sourceShardNode,
@ -1151,18 +1183,25 @@ CreateSplitShardsForShardGroupTwo(WorkerNode *sourceShardNode,
splitShardCreationCommandList,
shardInterval->shardId);
/* Create new split child shard on the specified placement list */
/* Create new split child shard on the specified worker node */
CreateObjectOnPlacement(splitShardCreationCommandList, workerPlacementNode);
}
}
}
static void
CreateDummyShardsForShardGroup(List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
WorkerNode *sourceWorkerNode,
List *workersForPlacementList)
{
/*
* Setup a hash map to store list of dummy shards created on nodes.
* This will facilitate easy cleanup.
*/
DummyShardInfoHashMap = SetupHashMapForShardInfo();
/*
* Statisfy Constraint 1: Create dummy source shard(s) on all destination nodes.
* If source node is also in desintation, skip dummy shard creation(see Note 1 from function description).
@ -1194,8 +1233,11 @@ CreateDummyShardsForShardGroup(List *sourceColocatedShardIntervalList,
splitShardCreationCommandList,
shardInterval->shardId);
/* Create new split child shard on the specified placement list */
/* Create dummy source shard on the specified placement list */
CreateObjectOnPlacement(splitShardCreationCommandList, workerPlacementNode);
/* Add dummy source shard entry created for placement node in map */
AddDummyShardEntryInMap(workerPlacementNode->nodeId, shardInterval);
}
}
@ -1224,12 +1266,16 @@ CreateDummyShardsForShardGroup(List *sourceColocatedShardIntervalList,
splitShardCreationCommandList,
shardInterval->shardId);
/* Create new split child shard on the specified placement list */
/* Create dummy split child shard on source worker node */
CreateObjectOnPlacement(splitShardCreationCommandList, sourceWorkerNode);
/* Add dummy split child shard entry created on source node */
AddDummyShardEntryInMap(sourceWorkerNode->nodeId, shardInterval);
}
}
}
static HTAB *
CreateWorkerForPlacementSet(List *workersForPlacementList)
{
@ -1257,13 +1303,14 @@ CreateWorkerForPlacementSet(List *workersForPlacementList)
}
static void SplitShardReplicationSetup(List *sourceColocatedShardIntervalList,
static void
SplitShardReplicationSetup(List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
WorkerNode *sourceWorkerNode,
List *destinationWorkerNodesList)
{
StringInfo splitShardReplicationUDF = CreateSplitShardReplicationSetupUDF(sourceColocatedShardIntervalList,
StringInfo splitShardReplicationUDF = CreateSplitShardReplicationSetupUDF(
sourceColocatedShardIntervalList,
shardGroupSplitIntervalListList,
destinationWorkerNodesList);
@ -1279,7 +1326,9 @@ static void SplitShardReplicationSetup(List *sourceColocatedShardIntervalList,
ClaimConnectionExclusively(sourceConnection);
PGresult *result = NULL;
int queryResult = ExecuteOptionalRemoteCommand(sourceConnection, splitShardReplicationUDF->data, &result);
int queryResult = ExecuteOptionalRemoteCommand(sourceConnection,
splitShardReplicationUDF->data,
&result);
if (queryResult != RESPONSE_OKAY || !IsResponseOK(result))
{
ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE),
@ -1292,12 +1341,89 @@ static void SplitShardReplicationSetup(List *sourceColocatedShardIntervalList,
/* Get replication slot information */
List *replicationSlotInfoList = ParseReplicationSlotInfoFromResult(result);
List * shardSplitPubSubMetadata = CreateShardSplitPubSubMetadataList(sourceColocatedShardIntervalList,
List *shardSplitPubSubMetadata = CreateShardSplitPubSubMetadataList(
sourceColocatedShardIntervalList,
shardGroupSplitIntervalListList,
destinationWorkerNodesList,
replicationSlotInfoList);
LogicallyReplicateSplitShards(sourceWorkerNode, shardSplitPubSubMetadata, sourceColocatedShardIntervalList, shardGroupSplitIntervalListList, destinationWorkerNodesList);
LogicallyReplicateSplitShards(sourceWorkerNode, shardSplitPubSubMetadata,
sourceColocatedShardIntervalList,
shardGroupSplitIntervalListList,
destinationWorkerNodesList);
}
static void
AddDummyShardEntryInMap(uint32 targetNodeId, ShardInterval *shardInterval)
{
NodeShardMappingKey key;
key.nodeId = targetNodeId;
key.tableOwnerId = TableOwnerOid(shardInterval->relationId);
bool found = false;
NodeShardMappingEntry *nodeMappingEntry =
(NodeShardMappingEntry *) hash_search(DummyShardInfoHashMap, &key, HASH_ENTER,
&found);
if (!found)
{
nodeMappingEntry->shardSplitInfoList = NIL;
}
nodeMappingEntry->shardSplitInfoList =
lappend(nodeMappingEntry->shardSplitInfoList, (ShardInterval *) shardInterval);
}
static void
DropDummyShards()
{
HASH_SEQ_STATUS status;
hash_seq_init(&status, DummyShardInfoHashMap);
NodeShardMappingEntry *entry = NULL;
while ((entry = (NodeShardMappingEntry *) hash_seq_search(&status)) != NULL)
{
uint32 nodeId = entry->key.nodeId;
WorkerNode *shardToBeDroppedNode = FindNodeWithNodeId(nodeId,
false /* missingOk */);
int connectionFlags = FOR_DDL;
connectionFlags |= OUTSIDE_TRANSACTION;
MultiConnection *connection = GetNodeUserDatabaseConnection(
connectionFlags,
shardToBeDroppedNode->workerName,
shardToBeDroppedNode->workerPort,
CurrentUserName(),
NULL /* databaseName */);
List *dummyShardIntervalList = entry->shardSplitInfoList;
ShardInterval *shardInterval = NULL;
foreach_ptr(shardInterval, dummyShardIntervalList)
{
TryDropShard(connection, shardInterval);
}
}
}
void
TryDropShard(MultiConnection *connection, ShardInterval *shardInterval)
{
char *qualifiedShardName = ConstructQualifiedShardName(shardInterval);
StringInfo dropShardQuery = makeStringInfo();
/* Caller enforces that foreign tables cannot be split (use DROP_REGULAR_TABLE_COMMAND) */
appendStringInfo(dropShardQuery, DROP_REGULAR_TABLE_COMMAND,
qualifiedShardName);
/*
* Perform a drop in best effort manner.
* The shard may or may not exist and the connection could have died.
*/
ExecuteOptionalRemoteCommand(
connection,
dropShardQuery->data,
NULL /* pgResult */);
}

View File

@ -48,7 +48,8 @@ static void AddShardSplitInfoEntryForNodeInMap(ShardSplitInfo *shardSplitInfo);
static void PopulateShardSplitInfoInSM(ShardSplitInfoSMHeader *shardSplitInfoSMHeader,
HTAB *shardInfoHashMap);
static void ReturnReplicationSlotInfo(HTAB *shardInfoHashMap, Tuplestorestate *tupleStore, TupleDesc tupleDescriptor);
static void ReturnReplicationSlotInfo(HTAB *shardInfoHashMap, Tuplestorestate *tupleStore,
TupleDesc tupleDescriptor);
static void CreatePublishersForSplitChildren(HTAB *shardInfoHashMap);
StringInfo GetSoureAndDestinationShardNames(List *shardSplitInfoList);
@ -420,9 +421,10 @@ ParseShardSplitInfoFromDatum(Datum shardSplitInfoDatum,
*nodeId = DatumGetInt32(nodeIdDatum);
}
static void CreatePublishersForSplitChildren(HTAB *shardInfoHashMap)
{
static void
CreatePublishersForSplitChildren(HTAB *shardInfoHashMap)
{
HASH_SEQ_STATUS status;
hash_seq_init(&status, shardInfoHashMap);
@ -441,16 +443,20 @@ static void CreatePublishersForSplitChildren(HTAB *shardInfoHashMap)
CitusExtensionOwnerName(),
get_database_name(
MyDatabaseId));
StringInfo shardNamesForPublication = GetSoureAndDestinationShardNames(entry->shardSplitInfoList);
StringInfo shardNamesForPublication = GetSoureAndDestinationShardNames(
entry->shardSplitInfoList);
StringInfo command = makeStringInfo();
appendStringInfo(command, "CREATE PUBLICATION sameerpub_%u_%u FOR TABLE %s", nodeId, tableOwnerId,shardNamesForPublication->data);
appendStringInfo(command, "CREATE PUBLICATION sameerpub_%u_%u FOR TABLE %s",
nodeId, tableOwnerId, shardNamesForPublication->data);
ExecuteCriticalRemoteCommand(sourceConnection, command->data);
printf("Sameer UserName: %s \n", GetUserNameFromId(tableOwnerId, false));
}
}
StringInfo GetSoureAndDestinationShardNames(List* shardSplitInfoList)
StringInfo
GetSoureAndDestinationShardNames(List *shardSplitInfoList)
{
HASHCTL info;
int flags = HASH_ELEM | HASH_CONTEXT;
@ -507,6 +513,7 @@ StringInfo GetSoureAndDestinationShardNames(List* shardSplitInfoList)
return allShardNames;
}
char *
ConstructFullyQualifiedSplitChildShardName(ShardSplitInfo *shardSplitInfo)
{
@ -521,7 +528,10 @@ ConstructFullyQualifiedSplitChildShardName(ShardSplitInfo* shardSplitInfo)
return shardName;
}
static void ReturnReplicationSlotInfo(HTAB *shardInfoHashMap, Tuplestorestate *tupleStore, TupleDesc tupleDescriptor)
static void
ReturnReplicationSlotInfo(HTAB *shardInfoHashMap, Tuplestorestate *tupleStore, TupleDesc
tupleDescriptor)
{
HASH_SEQ_STATUS status;
hash_seq_init(&status, shardInfoHashMap);
@ -540,7 +550,8 @@ static void ReturnReplicationSlotInfo(HTAB *shardInfoHashMap, Tuplestorestate *t
char *tableOwnerName = GetUserNameFromId(entry->key.tableOwnerId, false);
values[1] = CStringGetTextDatum(tableOwnerName);
char * slotName = encode_replication_slot(entry->key.nodeId, entry->key.tableOwnerId);
char *slotName = encode_replication_slot(entry->key.nodeId,
entry->key.tableOwnerId);
values[2] = CStringGetTextDatum(slotName);
tuplestore_putvalues(tupleStore, tupleDescriptor, values, nulls);

View File

@ -136,10 +136,12 @@ static XLogRecPtr GetRemoteLSN(MultiConnection *connection, char *command);
static uint64 TotalRelationSizeForSubscription(MultiConnection *connection,
char *command);
static bool RelationSubscriptionsAreReady(MultiConnection *targetConnection,
Bitmapset *tableOwnerIds, char * operationPrefix);
Bitmapset *tableOwnerIds,
char *operationPrefix);
static void WaitForMiliseconds(long timeout);
static XLogRecPtr GetSubscriptionPosition(MultiConnection *connection,
Bitmapset *tableOwnerIds, char * operationPrefix);
Bitmapset *tableOwnerIds,
char *operationPrefix);
static char * ShardMovePublicationName(Oid ownerId);
static char * ShardSubscriptionName(Oid ownerId, char *operationPrefix);
static void AcquireLogicalReplicationLock(void);
@ -148,7 +150,8 @@ static void DropAllShardMoveSubscriptions(MultiConnection *connection);
static void DropAllShardMoveReplicationSlots(MultiConnection *connection);
static void DropAllShardMovePublications(MultiConnection *connection);
static void DropAllShardMoveUsers(MultiConnection *connection);
static char * ShardSubscriptionNamesValueList(Bitmapset *tableOwnerIds, char * operationPrefix);
static char * ShardSubscriptionNamesValueList(Bitmapset *tableOwnerIds,
char *operationPrefix);
static void DropShardMoveSubscription(MultiConnection *connection,
char *subscriptionName);
static void DropShardMoveReplicationSlot(MultiConnection *connection,
@ -227,14 +230,16 @@ LogicallyReplicateShards(List *shardList, char *sourceNodeName, int sourceNodePo
* subscription is not ready. There is no point of locking the shards before the
* subscriptions for each relation becomes ready, so wait for it.
*/
WaitForRelationSubscriptionsBecomeReady(targetConnection, tableOwnerIds, SHARD_MOVE_SUBSCRIPTION_PREFIX);
WaitForRelationSubscriptionsBecomeReady(targetConnection, tableOwnerIds,
SHARD_MOVE_SUBSCRIPTION_PREFIX);
/*
* Wait until the subscription is caught up to changes that has happened
* after the initial COPY on the shards.
*/
XLogRecPtr sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds, SHARD_MOVE_SUBSCRIPTION_PREFIX);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds,
SHARD_MOVE_SUBSCRIPTION_PREFIX);
/*
* Now lets create the post-load objects, such as the indexes, constraints
@ -244,7 +249,8 @@ LogicallyReplicateShards(List *shardList, char *sourceNodeName, int sourceNodePo
CreatePostLogicalReplicationDataLoadObjects(shardList, targetNodeName,
targetNodePort);
sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds, SHARD_MOVE_SUBSCRIPTION_PREFIX);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds,
SHARD_MOVE_SUBSCRIPTION_PREFIX);
/*
* We're almost done, we'll block the writes to the shards that we're
@ -257,7 +263,8 @@ LogicallyReplicateShards(List *shardList, char *sourceNodeName, int sourceNodePo
BlockWritesToShardList(shardList);
sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds, SHARD_MOVE_SUBSCRIPTION_PREFIX);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds,
SHARD_MOVE_SUBSCRIPTION_PREFIX);
/*
* We're creating the foreign constraints to reference tables after the
@ -1085,7 +1092,8 @@ DropShardMovePublications(MultiConnection *connection, Bitmapset *tableOwnerIds)
* If replication slot can not be dropped while dropping the subscriber, drop
* it here.
*/
DropShardMoveReplicationSlot(connection, ShardSubscriptionName(ownerId, SHARD_MOVE_SUBSCRIPTION_PREFIX));
DropShardMoveReplicationSlot(connection, ShardSubscriptionName(ownerId,
SHARD_MOVE_SUBSCRIPTION_PREFIX));
DropShardMovePublication(connection, ShardMovePublicationName(ownerId));
}
}
@ -1317,7 +1325,8 @@ DropShardMoveSubscriptions(MultiConnection *connection, Bitmapset *tableOwnerIds
int ownerId = -1;
while ((ownerId = bms_next_member(tableOwnerIds, ownerId)) >= 0)
{
DropShardMoveSubscription(connection, ShardSubscriptionName(ownerId, SHARD_MOVE_SUBSCRIPTION_PREFIX));
DropShardMoveSubscription(connection, ShardSubscriptionName(ownerId,
SHARD_MOVE_SUBSCRIPTION_PREFIX));
DropShardMoveUser(connection, ShardMoveSubscriptionRole(ownerId));
}
}
@ -1491,7 +1500,8 @@ CreateShardMoveSubscriptions(MultiConnection *connection, char *sourceNodeName,
appendStringInfo(createSubscriptionCommand,
"CREATE SUBSCRIPTION %s CONNECTION %s PUBLICATION %s "
"WITH (citus_use_authinfo=true, enabled=false)",
quote_identifier(ShardSubscriptionName(ownerId, SHARD_MOVE_SUBSCRIPTION_PREFIX)),
quote_identifier(ShardSubscriptionName(ownerId,
SHARD_MOVE_SUBSCRIPTION_PREFIX)),
quote_literal_cstr(conninfo->data),
quote_identifier(ShardMovePublicationName(ownerId)));
@ -1500,7 +1510,8 @@ CreateShardMoveSubscriptions(MultiConnection *connection, char *sourceNodeName,
pfree(createSubscriptionCommand);
ExecuteCriticalRemoteCommand(connection, psprintf(
"ALTER SUBSCRIPTION %s OWNER TO %s",
ShardSubscriptionName(ownerId, SHARD_MOVE_SUBSCRIPTION_PREFIX),
ShardSubscriptionName(ownerId,
SHARD_MOVE_SUBSCRIPTION_PREFIX),
ShardMoveSubscriptionRole(ownerId)
));
@ -1519,7 +1530,8 @@ CreateShardMoveSubscriptions(MultiConnection *connection, char *sourceNodeName,
ExecuteCriticalRemoteCommand(connection, psprintf(
"ALTER SUBSCRIPTION %s ENABLE",
ShardSubscriptionName(ownerId, SHARD_MOVE_SUBSCRIPTION_PREFIX)
ShardSubscriptionName(ownerId,
SHARD_MOVE_SUBSCRIPTION_PREFIX)
));
}
}
@ -1655,7 +1667,8 @@ WaitForRelationSubscriptionsBecomeReady(MultiConnection *targetConnection,
while (true)
{
/* we're done, all relations are ready */
if (RelationSubscriptionsAreReady(targetConnection, tableOwnerIds, operationPrefix))
if (RelationSubscriptionsAreReady(targetConnection, tableOwnerIds,
operationPrefix))
{
ereport(LOG, (errmsg("The states of the relations belonging to the "
"subscriptions became READY on the "
@ -1665,7 +1678,8 @@ WaitForRelationSubscriptionsBecomeReady(MultiConnection *targetConnection,
break;
}
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds, operationPrefix);
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds,
operationPrefix);
/* Get the current total size of tables belonging to the subscriber */
uint64 currentTotalRelationSize =
@ -1842,7 +1856,8 @@ ShardSubscriptionNamesValueList(Bitmapset *tableOwnerIds, char * operationPrefix
first = false;
}
appendStringInfoString(subscriptionValueList,
quote_literal_cstr(ShardSubscriptionName(ownerId, operationPrefix)));
quote_literal_cstr(ShardSubscriptionName(ownerId,
operationPrefix)));
}
appendStringInfoString(subscriptionValueList, ")");
return subscriptionValueList->data;
@ -1859,7 +1874,8 @@ RelationSubscriptionsAreReady(MultiConnection *targetConnection,
{
bool raiseInterrupts = false;
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds, operationPrefix);
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds,
operationPrefix);
char *query = psprintf(
"SELECT count(*) FROM pg_subscription_rel, pg_stat_subscription "
"WHERE srsubid = subid AND srsubstate != 'r' AND subname IN %s",
@ -1910,7 +1926,8 @@ RelationSubscriptionsAreReady(MultiConnection *targetConnection,
* The function also reports its progress in every logicalReplicationProgressReportTimeout.
*/
void
WaitForShardSubscriptionToCatchUp(MultiConnection *targetConnection, XLogRecPtr sourcePosition,
WaitForShardSubscriptionToCatchUp(MultiConnection *targetConnection, XLogRecPtr
sourcePosition,
Bitmapset *tableOwnerIds, char *operationPrefix)
{
XLogRecPtr previousTargetPosition = 0;
@ -2050,9 +2067,11 @@ WaitForMiliseconds(long timeout)
* replication.
*/
static XLogRecPtr
GetSubscriptionPosition(MultiConnection *connection, Bitmapset *tableOwnerIds, char * operationPrefix)
GetSubscriptionPosition(MultiConnection *connection, Bitmapset *tableOwnerIds,
char *operationPrefix)
{
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds, operationPrefix);
char *subscriptionValueList = ShardSubscriptionNamesValueList(tableOwnerIds,
operationPrefix);
return GetRemoteLSN(connection, psprintf(
"SELECT min(latest_end_lsn) FROM pg_stat_subscription "
"WHERE subname IN %s", subscriptionValueList));
@ -2066,7 +2085,6 @@ CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
char *publicationName,
Oid ownerId)
{
StringInfo createSubscriptionCommand = makeStringInfo();
StringInfo conninfo = makeStringInfo();
@ -2092,7 +2110,8 @@ CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
appendStringInfo(createSubscriptionCommand,
"CREATE SUBSCRIPTION %s CONNECTION %s PUBLICATION %s "
"WITH (citus_use_authinfo=true, enabled=false)",
quote_identifier(ShardSubscriptionName(ownerId, SHARD_SPLIT_SUBSCRIPTION_PREFIX)),
quote_identifier(ShardSubscriptionName(ownerId,
SHARD_SPLIT_SUBSCRIPTION_PREFIX)),
quote_literal_cstr(conninfo->data),
quote_identifier(publicationName));
@ -2101,7 +2120,8 @@ CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
pfree(createSubscriptionCommand);
ExecuteCriticalRemoteCommand(connection, psprintf(
"ALTER SUBSCRIPTION %s OWNER TO %s",
ShardSubscriptionName(ownerId, SHARD_SPLIT_SUBSCRIPTION_PREFIX),
ShardSubscriptionName(ownerId,
SHARD_SPLIT_SUBSCRIPTION_PREFIX),
ShardMoveSubscriptionRole(ownerId)
));
@ -2120,6 +2140,7 @@ CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
ExecuteCriticalRemoteCommand(connection, psprintf(
"ALTER SUBSCRIPTION %s ENABLE",
ShardSubscriptionName(ownerId, SHARD_SPLIT_SUBSCRIPTION_PREFIX)
ShardSubscriptionName(ownerId,
SHARD_SPLIT_SUBSCRIPTION_PREFIX)
));
}

View File

@ -26,30 +26,39 @@
static HTAB *ShardInfoHashMapForPublications = NULL;
/* function declarations */
static void AddShardEntryInMap(Oid tableOwner, uint32 nodeId, ShardInterval * shardInterval, bool isChildShardInterval);
ShardSplitPubSubMetadata * CreateShardSplitPubSubMetadata(Oid tableOwnerId, uint32 nodeId, List * shardIdList, List * replicationSlotInfoList);
static void AddPublishableShardEntryInMap(uint32 targetNodeId,
ShardInterval *shardInterval, bool
isChildShardInterval);
ShardSplitPubSubMetadata * CreateShardSplitPubSubMetadata(Oid tableOwnerId, uint32 nodeId,
List *shardIdList,
List *replicationSlotInfoList);
static void
CreateShardSplitPublicationForNode(MultiConnection *connection, List *shardList,
uint32_t publicationForTargetNodeId, Oid tableOwner);
static void
CreateShardSplitPublications(MultiConnection *sourceConnection, List * shardSplitPubSubMetadataList);
static void
CreateShardSplitSubscriptions(List * targetNodeConnectionList, List * shardSplitPubSubMetadataList, WorkerNode * sourceWorkerNode, char * superUser, char * databaseName);
static void
WaitForShardSplitRelationSubscriptionsBecomeReady(List * targetNodeConnectionList, List * shardSplitPubSubMetadataList);
static void CreateShardSplitPublicationForNode(MultiConnection *connection,
List *shardList,
uint32_t publicationForTargetNodeId, Oid
tableOwner);
static void CreateShardSplitPublications(MultiConnection *sourceConnection,
List *shardSplitPubSubMetadataList);
static void CreateShardSplitSubscriptions(List *targetNodeConnectionList,
List *shardSplitPubSubMetadataList,
WorkerNode *sourceWorkerNode, char *superUser,
char *databaseName);
static void WaitForShardSplitRelationSubscriptionsBecomeReady(
List *targetNodeConnectionList, List *shardSplitPubSubMetadataList);
static void
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(XLogRecPtr sourcePosition, List * targetNodeConnectionList, List * shardSplitPubSubMetadataList);
static void WaitForShardSplitRelationSubscriptionsToBeCaughtUp(XLogRecPtr sourcePosition,
List *
targetNodeConnectionList,
List *
shardSplitPubSubMetadataList);
static char *
ShardSplitPublicationName(uint32_t nodeId, Oid ownerId);
static char * ShardSplitPublicationName(uint32_t nodeId, Oid ownerId);
/*used for debuggin. Remove later*/
void PrintShardSplitPubSubMetadata(ShardSplitPubSubMetadata *shardSplitMetadata);
StringInfo CreateSplitShardReplicationSetupUDF(List *sourceColocatedShardIntervalList,
StringInfo
CreateSplitShardReplicationSetupUDF(List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
List *destinationWorkerNodesList)
{
@ -74,10 +83,12 @@ StringInfo CreateSplitShardReplicationSetupUDF(List *sourceColocatedShardInterva
}
StringInfo minValueString = makeStringInfo();
appendStringInfo(minValueString, "%d", DatumGetInt32(splitChildShardInterval->minValue));
appendStringInfo(minValueString, "%d", DatumGetInt32(
splitChildShardInterval->minValue));
StringInfo maxValueString = makeStringInfo();
appendStringInfo(maxValueString, "%d", DatumGetInt32(splitChildShardInterval->maxValue));
appendStringInfo(maxValueString, "%d", DatumGetInt32(
splitChildShardInterval->maxValue));
appendStringInfo(splitChildrenRows,
"ROW(%lu, %lu, %s, %s, %u)::citus.split_shard_info",
@ -93,12 +104,15 @@ StringInfo CreateSplitShardReplicationSetupUDF(List *sourceColocatedShardInterva
StringInfo splitShardReplicationUDF = makeStringInfo();
appendStringInfo(splitShardReplicationUDF,
"SELECT * FROM worker_split_shard_replication_setup(ARRAY[%s])", splitChildrenRows->data);
"SELECT * FROM worker_split_shard_replication_setup(ARRAY[%s])",
splitChildrenRows->data);
return splitShardReplicationUDF;
}
List * ParseReplicationSlotInfoFromResult(PGresult * result)
List *
ParseReplicationSlotInfoFromResult(PGresult *result)
{
int64 rowCount = PQntuples(result);
int64 colCount = PQnfields(result);
@ -106,7 +120,8 @@ List * ParseReplicationSlotInfoFromResult(PGresult * result)
List *replicationSlotInfoList = NIL;
for (int64 rowIndex = 0; rowIndex < rowCount; rowIndex++)
{
ReplicationSlotInfo * replicationSlotInfo = (ReplicationSlotInfo *)palloc0(sizeof(ReplicationSlotInfo));
ReplicationSlotInfo *replicationSlotInfo = (ReplicationSlotInfo *) palloc0(
sizeof(ReplicationSlotInfo));
char *targeNodeIdString = PQgetvalue(result, rowIndex, 0);
replicationSlotInfo->targetNodeId = strtoul(targeNodeIdString, NULL, 10);
@ -124,7 +139,8 @@ List * ParseReplicationSlotInfoFromResult(PGresult * result)
}
List * CreateShardSplitPubSubMetadataList(List *sourceColocatedShardIntervalList,
List *
CreateShardSplitPubSubMetadataList(List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
List *destinationWorkerNodesList,
List *replicationSlotInfoList)
@ -140,15 +156,17 @@ List * CreateShardSplitPubSubMetadataList(List *sourceColocatedShardIntervalList
forboth_ptr(splitChildShardInterval, splitChildShardIntervalList,
destinationWorkerNode, destinationWorkerNodesList)
{
/* Table owner is same for both parent and child shard */
Oid tableOwnerId = TableOwnerOid(sourceShardIntervalToCopy->relationId);
uint32 destinationWorkerNodeId = destinationWorkerNode->nodeId;
/* Add split child shard interval */
AddShardEntryInMap(tableOwnerId, destinationWorkerNodeId, splitChildShardInterval, true /*isChildShardInterval*/);
AddPublishableShardEntryInMap(destinationWorkerNodeId,
splitChildShardInterval,
true /*isChildShardInterval*/);
/* Add parent shard interval if not already added */
AddShardEntryInMap(tableOwnerId, destinationWorkerNodeId, sourceShardIntervalToCopy, false /*isChildShardInterval*/);
AddPublishableShardEntryInMap(destinationWorkerNodeId,
sourceShardIntervalToCopy,
false /*isChildShardInterval*/);
}
}
@ -162,23 +180,31 @@ List * CreateShardSplitPubSubMetadataList(List *sourceColocatedShardIntervalList
{
uint32_t nodeId = entry->key.nodeId;
uint32_t tableOwnerId = entry->key.tableOwnerId;
ShardSplitPubSubMetadata * shardSplitPubSubMetadata = CreateShardSplitPubSubMetadata(tableOwnerId, nodeId, entry->shardSplitInfoList, replicationSlotInfoList);
ShardSplitPubSubMetadata *shardSplitPubSubMetadata =
CreateShardSplitPubSubMetadata(tableOwnerId, nodeId,
entry->shardSplitInfoList,
replicationSlotInfoList);
shardSplitPubSubMetadataList = lappend(shardSplitPubSubMetadataList, shardSplitPubSubMetadata);
shardSplitPubSubMetadataList = lappend(shardSplitPubSubMetadataList,
shardSplitPubSubMetadata);
}
return shardSplitPubSubMetadataList;
}
static void AddShardEntryInMap(Oid tableOwnerId, uint32 nodeId, ShardInterval * shardInterval, bool isChildShardInterval)
static void
AddPublishableShardEntryInMap(uint32 targetNodeId, ShardInterval *shardInterval, bool
isChildShardInterval)
{
NodeShardMappingKey key;
key.nodeId = nodeId;
key.tableOwnerId = tableOwnerId;
key.nodeId = targetNodeId;
key.tableOwnerId = TableOwnerOid(shardInterval->relationId);
bool found = false;
NodeShardMappingEntry *nodeMappingEntry =
(NodeShardMappingEntry *) hash_search(ShardInfoHashMapForPublications, &key, HASH_ENTER,
(NodeShardMappingEntry *) hash_search(ShardInfoHashMapForPublications, &key,
HASH_ENTER,
&found);
if (!found)
{
@ -188,7 +214,8 @@ static void AddShardEntryInMap(Oid tableOwnerId, uint32 nodeId, ShardInterval *
if (isChildShardInterval)
{
nodeMappingEntry->shardSplitInfoList =
lappend(nodeMappingEntry->shardSplitInfoList, (ShardInterval *) shardInterval);
lappend(nodeMappingEntry->shardSplitInfoList,
(ShardInterval *) shardInterval);
return;
}
@ -205,14 +232,15 @@ static void AddShardEntryInMap(Oid tableOwnerId, uint32 nodeId, ShardInterval *
/* Add parent shard Interval */
nodeMappingEntry->shardSplitInfoList =
lappend(nodeMappingEntry->shardSplitInfoList, (ShardInterval *) shardInterval);
}
ShardSplitPubSubMetadata * CreateShardSplitPubSubMetadata(Oid tableOwnerId, uint32 nodeId, List * shardIntervalList, List * replicationSlotInfoList)
ShardSplitPubSubMetadata *
CreateShardSplitPubSubMetadata(Oid tableOwnerId, uint32 nodeId, List *shardIntervalList,
List *replicationSlotInfoList)
{
ShardSplitPubSubMetadata * shardSplitPubSubMetadata = palloc0(sizeof(ShardSplitPubSubMetadata));
ShardSplitPubSubMetadata *shardSplitPubSubMetadata = palloc0(
sizeof(ShardSplitPubSubMetadata));
shardSplitPubSubMetadata->shardIntervalListForSubscription = shardIntervalList;
shardSplitPubSubMetadata->tableOwnerId = tableOwnerId;
@ -232,7 +260,8 @@ ShardSplitPubSubMetadata * CreateShardSplitPubSubMetadata(Oid tableOwnerId, uint
}
void LogicallyReplicateSplitShards(WorkerNode *sourceWorkerNode,
void
LogicallyReplicateSplitShards(WorkerNode *sourceWorkerNode,
List *shardSplitPubSubMetadataList,
List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
@ -244,7 +273,8 @@ void LogicallyReplicateSplitShards(WorkerNode *sourceWorkerNode,
/* Get source node connection */
MultiConnection *sourceConnection =
GetNodeUserDatabaseConnection(connectionFlags, sourceWorkerNode->workerName, sourceWorkerNode->workerPort,
GetNodeUserDatabaseConnection(connectionFlags, sourceWorkerNode->workerName,
sourceWorkerNode->workerPort,
superUser, databaseName);
ClaimConnectionExclusively(sourceConnection);
@ -257,7 +287,8 @@ void LogicallyReplicateSplitShards(WorkerNode *sourceWorkerNode,
WorkerNode *targetWorkerNode = FindNodeWithNodeId(targetWorkerNodeId, false);
MultiConnection *targetConnection =
GetNodeUserDatabaseConnection(connectionFlags, targetWorkerNode->workerName, targetWorkerNode->workerPort,
GetNodeUserDatabaseConnection(connectionFlags, targetWorkerNode->workerName,
targetWorkerNode->workerPort,
superUser, databaseName);
ClaimConnectionExclusively(targetConnection);
@ -273,26 +304,36 @@ void LogicallyReplicateSplitShards(WorkerNode *sourceWorkerNode,
superUser,
databaseName);
WaitForShardSplitRelationSubscriptionsBecomeReady(targetNodeConnectionList, shardSplitPubSubMetadataList);
WaitForShardSplitRelationSubscriptionsBecomeReady(targetNodeConnectionList,
shardSplitPubSubMetadataList);
XLogRecPtr sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition, targetNodeConnectionList, shardSplitPubSubMetadataList);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition,
targetNodeConnectionList,
shardSplitPubSubMetadataList);
CreateAuxiliaryStructuresForShardGroup(shardGroupSplitIntervalListList, destinationWorkerNodesList);
CreateAuxiliaryStructuresForShardGroup(shardGroupSplitIntervalListList,
destinationWorkerNodesList);
sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition, targetNodeConnectionList, shardSplitPubSubMetadataList);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition,
targetNodeConnectionList,
shardSplitPubSubMetadataList);
BlockWritesToShardList(sourceColocatedShardIntervalList);
sourcePosition = GetRemoteLogPosition(sourceConnection);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition, targetNodeConnectionList, shardSplitPubSubMetadataList);
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(sourcePosition,
targetNodeConnectionList,
shardSplitPubSubMetadataList);
/*TOOD : Create foreign key constraints and handle partitioned tables*/
}
void PrintShardSplitPubSubMetadata(ShardSplitPubSubMetadata * shardSplitMetadata)
void
PrintShardSplitPubSubMetadata(ShardSplitPubSubMetadata *shardSplitMetadata)
{
printf("sameer: ShardSplitPubSbuMetadata\n");
ReplicationSlotInfo *replicationInfo = shardSplitMetadata->slotInfo;
@ -305,13 +346,16 @@ void PrintShardSplitPubSubMetadata(ShardSplitPubSubMetadata * shardSplitMetadata
printf("%ld ", shardInterval->shardId);
}
printf("\nManual Username from OID at source: %s \n", GetUserNameFromId(shardSplitMetadata->tableOwnerId, false));
printf("slotname:%s targetNode:%u tableOwner:%s \n", replicationInfo->slotName, replicationInfo->targetNodeId, replicationInfo->tableOwnerName);
printf("\nManual Username from OID at source: %s \n", GetUserNameFromId(
shardSplitMetadata->tableOwnerId, false));
printf("slotname:%s targetNode:%u tableOwner:%s \n", replicationInfo->slotName,
replicationInfo->targetNodeId, replicationInfo->tableOwnerName);
}
static void
CreateShardSplitPublications(MultiConnection *sourceConnection, List *shardSplitPubSubMetadataList)
CreateShardSplitPublications(MultiConnection *sourceConnection,
List *shardSplitPubSubMetadataList)
{
ShardSplitPubSubMetadata *shardSplitPubSubMetadata = NULL;
foreach_ptr(shardSplitPubSubMetadata, shardSplitPubSubMetadataList)
@ -320,12 +364,14 @@ CreateShardSplitPublications(MultiConnection *sourceConnection, List *shardSplit
Oid tableOwnerId = shardSplitPubSubMetadata->tableOwnerId;
CreateShardSplitPublicationForNode(sourceConnection,
shardSplitPubSubMetadata->shardIntervalListForSubscription,
shardSplitPubSubMetadata->
shardIntervalListForSubscription,
publicationForNodeId,
tableOwnerId);
}
}
static void
CreateShardSplitSubscriptions(List *targetNodeConnectionList,
List *shardSplitPubSubMetadataList,
@ -355,7 +401,6 @@ static void
CreateShardSplitPublicationForNode(MultiConnection *connection, List *shardList,
uint32_t publicationForTargetNodeId, Oid ownerId)
{
StringInfo createPublicationCommand = makeStringInfo();
bool prefixWithComma = false;
@ -390,7 +435,8 @@ ShardSplitPublicationName(uint32_t nodeId, Oid ownerId)
static void
WaitForShardSplitRelationSubscriptionsBecomeReady(List * targetNodeConnectionList, List * shardSplitPubSubMetadataList)
WaitForShardSplitRelationSubscriptionsBecomeReady(List *targetNodeConnectionList,
List *shardSplitPubSubMetadataList)
{
MultiConnection *targetConnection = NULL;
ShardSplitPubSubMetadata *shardSplitPubSubMetadata = NULL;
@ -398,13 +444,18 @@ WaitForShardSplitRelationSubscriptionsBecomeReady(List * targetNodeConnectionLis
shardSplitPubSubMetadata, shardSplitPubSubMetadataList)
{
Bitmapset *tableOwnerIds = NULL;
tableOwnerIds = bms_add_member(tableOwnerIds, shardSplitPubSubMetadata->tableOwnerId);
WaitForRelationSubscriptionsBecomeReady(targetConnection, tableOwnerIds, SHARD_SPLIT_SUBSCRIPTION_PREFIX);
tableOwnerIds = bms_add_member(tableOwnerIds,
shardSplitPubSubMetadata->tableOwnerId);
WaitForRelationSubscriptionsBecomeReady(targetConnection, tableOwnerIds,
SHARD_SPLIT_SUBSCRIPTION_PREFIX);
}
}
static void
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(XLogRecPtr sourcePosition, List * targetNodeConnectionList, List * shardSplitPubSubMetadataList)
WaitForShardSplitRelationSubscriptionsToBeCaughtUp(XLogRecPtr sourcePosition,
List *targetNodeConnectionList,
List *shardSplitPubSubMetadataList)
{
MultiConnection *targetConnection = NULL;
ShardSplitPubSubMetadata *shardSplitPubSubMetadata = NULL;
@ -412,7 +463,9 @@ WaitForShardSplitRelationSubscriptionsToBeCaughtUp(XLogRecPtr sourcePosition, Li
shardSplitPubSubMetadata, shardSplitPubSubMetadataList)
{
Bitmapset *tableOwnerIds = NULL;
tableOwnerIds = bms_add_member(tableOwnerIds, shardSplitPubSubMetadata->tableOwnerId);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds, SHARD_SPLIT_SUBSCRIPTION_PREFIX);
tableOwnerIds = bms_add_member(tableOwnerIds,
shardSplitPubSubMetadata->tableOwnerId);
WaitForShardSubscriptionToCatchUp(targetConnection, sourcePosition, tableOwnerIds,
SHARD_SPLIT_SUBSCRIPTION_PREFIX);
}
}

View File

@ -28,14 +28,15 @@ extern void LogicallyReplicateShards(List *shardList, char *sourceNodeName,
extern XLogRecPtr GetRemoteLogPosition(MultiConnection *connection);
extern void
CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
int sourceNodePort, char *userName, char *databaseName,
extern void CreateShardSubscription(MultiConnection *connection, char *sourceNodeName,
int sourceNodePort, char *userName,
char *databaseName,
char *publicationName,
Oid ownerId);
extern void WaitForRelationSubscriptionsBecomeReady(MultiConnection *targetConnection,
Bitmapset *tableOwnerIds, char * operationPrefix);
Bitmapset *tableOwnerIds,
char *operationPrefix);
extern void WaitForShardSubscriptionToCatchUp(MultiConnection *targetConnection,
XLogRecPtr sourcePosition,
Bitmapset *tableOwnerIds,

View File

@ -44,8 +44,10 @@ HTAB * SetupHashMapForShardInfo(void);
List * ParseReplicationSlotInfoFromResult(PGresult *result);
extern StringInfo CreateSplitShardReplicationSetupUDF(List *sourceColocatedShardIntervalList,
List *shardGroupSplitIntervalListList,
extern StringInfo CreateSplitShardReplicationSetupUDF(
List *sourceColocatedShardIntervalList,
List *
shardGroupSplitIntervalListList,
List *destinationWorkerNodesList);
extern List * CreateShardSplitPubSubMetadataList(List *sourceColocatedShardIntervalList,

View File

@ -113,8 +113,8 @@ SELECT nodeid AS worker_2_node FROM pg_dist_node WHERE nodeport=:worker_2_port \
SELECT citus_split_shard_by_split_points(
1,
ARRAY['-1073741826'],
ARRAY[:worker_1_node, :worker_2_node],
'non_blocking');
ARRAY[:worker_2_node, :worker_2_node],
'force_logical');
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (100, 'CREATE TABLE citus_split_shard_by_split_points_negative.table_to_split (id bigint NOT NULL, value character(1)) ')
@ -165,6 +165,14 @@ NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (100, 'CREATE TABLE citus_split_shard_by_split_points_negative.table_to_split (id bigint NOT NULL, value character(1)) ')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (100, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_to_split OWNER TO postgres')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (101, 'CREATE TABLE citus_split_shard_by_split_points_negative.table_to_split (id bigint NOT NULL, value character(1)) ')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (101, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_to_split OWNER TO postgres')
@ -173,14 +181,121 @@ NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (102, 'CREATE TABLE citus_split_shard_by_split_points_negative.table_second (id bigint NOT NULL, value character(1)) ')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (102, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_second OWNER TO postgres')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (103, 'CREATE TABLE citus_split_shard_by_split_points_negative.table_second (id bigint NOT NULL, value character(1)) ')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (103, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_second OWNER TO postgres')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT * FROM worker_split_shard_replication_setup(ARRAY[ROW(1, 100, '-2147483648', '-1073741826', 16)::citus.split_shard_info,ROW(1, 101, '-1073741825', '2147483647', 18)::citus.split_shard_info,ROW(2, 102, '-2147483648', '-1073741826', 16)::citus.split_shard_info,ROW(2, 103, '-1073741825', '2147483647', 18)::citus.split_shard_info])
NOTICE: issuing SELECT * FROM worker_split_shard_replication_setup(ARRAY[ROW(1, 100, '-2147483648', '-1073741826', 18)::citus.split_shard_info,ROW(1, 101, '-1073741825', '2147483647', 18)::citus.split_shard_info,ROW(2, 102, '-2147483648', '-1073741826', 18)::citus.split_shard_info,ROW(2, 103, '-1073741825', '2147483647', 18)::citus.split_shard_info])
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing CREATE PUBLICATION citus_shard_split_publication_18_10 FOR TABLE citus_split_shard_by_split_points_negative.table_to_split_100,citus_split_shard_by_split_points_negative.table_to_split_1,citus_split_shard_by_split_points_negative.table_to_split_101,citus_split_shard_by_split_points_negative.table_second_102,citus_split_shard_by_split_points_negative.table_second_2,citus_split_shard_by_split_points_negative.table_second_103
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SET LOCAL citus.enable_ddl_propagation TO OFF;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing CREATE USER citus_shard_move_subscription_role_10 SUPERUSER IN ROLE postgres
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing CREATE SUBSCRIPTION citus_shard_split_subscription_10 CONNECTION 'host=''localhost'' port=xxxxx user=''postgres'' dbname=''regression'' connect_timeout=20' PUBLICATION citus_shard_split_publication_18_10 WITH (citus_use_authinfo=true, enabled=false)
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing ALTER SUBSCRIPTION citus_shard_split_subscription_10 OWNER TO citus_shard_move_subscription_role_10
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SET LOCAL citus.enable_ddl_propagation TO OFF;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing ALTER ROLE citus_shard_move_subscription_role_10 NOSUPERUSER
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing ALTER SUBSCRIPTION citus_shard_split_subscription_10 ENABLE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT count(*) FROM pg_subscription_rel, pg_stat_subscription WHERE srsubid = subid AND srsubstate != 'r' AND subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT sum(pg_total_relation_size(srrelid)) FROM pg_subscription_rel, pg_stat_subscription WHERE srsubid = subid AND subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT count(*) FROM pg_subscription_rel, pg_stat_subscription WHERE srsubid = subid AND srsubstate != 'r' AND subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT pg_current_wal_lsn()
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT min(latest_end_lsn) FROM pg_stat_subscription WHERE subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (100, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_to_split ADD CONSTRAINT table_to_split_pkey PRIMARY KEY (id)')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (101, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_to_split ADD CONSTRAINT table_to_split_pkey PRIMARY KEY (id)')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (102, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_second ADD CONSTRAINT table_second_pkey PRIMARY KEY (id)')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT worker_apply_shard_ddl_command (103, 'ALTER TABLE citus_split_shard_by_split_points_negative.table_second ADD CONSTRAINT table_second_pkey PRIMARY KEY (id)')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT pg_current_wal_lsn()
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT min(latest_end_lsn) FROM pg_stat_subscription WHERE subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing BEGIN TRANSACTION ISOLATION LEVEL READ COMMITTED;SELECT assign_distributed_transaction_id(xx, xx, 'xxxxxxx');
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT lock_shard_metadata(7, ARRAY[1, 2])
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT lock_shard_metadata(7, ARRAY[1, 2])
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT pg_current_wal_lsn()
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT min(latest_end_lsn) FROM pg_stat_subscription WHERE subname IN ('citus_shard_split_subscription_10')
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT citus_internal_delete_shard_metadata(1);
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT citus_internal_delete_shard_metadata(1);
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_to_split_1 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT citus_internal_delete_shard_metadata(2);
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing SELECT citus_internal_delete_shard_metadata(2);
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_second_2 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing WITH shard_data(relationname, shardid, storagetype, shardminvalue, shardmaxvalue) AS (VALUES ('citus_split_shard_by_split_points_negative.table_to_split'::regclass, 100, 't'::"char", '-2147483648', '-1073741826'), ('citus_split_shard_by_split_points_negative.table_to_split'::regclass, 101, 't'::"char", '-1073741825', '2147483647'), ('citus_split_shard_by_split_points_negative.table_second'::regclass, 102, 't'::"char", '-2147483648', '-1073741826'), ('citus_split_shard_by_split_points_negative.table_second'::regclass, 103, 't'::"char", '-1073741825', '2147483647')) SELECT citus_internal_add_shard_metadata(relationname, shardid, storagetype, shardminvalue, shardmaxvalue) FROM shard_data;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing WITH shard_data(relationname, shardid, storagetype, shardminvalue, shardmaxvalue) AS (VALUES ('citus_split_shard_by_split_points_negative.table_to_split'::regclass, 100, 't'::"char", '-2147483648', '-1073741826'), ('citus_split_shard_by_split_points_negative.table_to_split'::regclass, 101, 't'::"char", '-1073741825', '2147483647'), ('citus_split_shard_by_split_points_negative.table_second'::regclass, 102, 't'::"char", '-2147483648', '-1073741826'), ('citus_split_shard_by_split_points_negative.table_second'::regclass, 103, 't'::"char", '-1073741825', '2147483647')) SELECT citus_internal_add_shard_metadata(relationname, shardid, storagetype, shardminvalue, shardmaxvalue) FROM shard_data;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing WITH placement_data(shardid, shardstate, shardlength, groupid, placementid) AS (VALUES (100, 1, 0, 16, 85), (101, 1, 0, 16, 86), (102, 1, 0, 16, 87), (103, 1, 0, 16, 88)) SELECT citus_internal_add_placement_metadata(shardid, shardstate, shardlength, groupid, placementid) FROM placement_data;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing WITH placement_data(shardid, shardstate, shardlength, groupid, placementid) AS (VALUES (100, 1, 0, 16, 85), (101, 1, 0, 16, 86), (102, 1, 0, 16, 87), (103, 1, 0, 16, 88)) SELECT citus_internal_add_placement_metadata(shardid, shardstate, shardlength, groupid, placementid) FROM placement_data;
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_to_split_100 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_to_split_101 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_second_102 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_second_103 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_to_split_1 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing DROP TABLE IF EXISTS citus_split_shard_by_split_points_negative.table_second_2 CASCADE
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing PREPARE TRANSACTION 'citus_xx_xx_xx_xx'
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing PREPARE TRANSACTION 'citus_xx_xx_xx_xx'
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT PREPARED 'citus_xx_xx_xx_xx'
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
NOTICE: issuing COMMIT PREPARED 'citus_xx_xx_xx_xx'
DETAIL: on server postgres@localhost:xxxxx connectionId: xxxxxxx
WARNING: connection claimed exclusively at transaction commit
WARNING: connection claimed exclusively at transaction commit
WARNING: connection claimed exclusively at transaction commit
citus_split_shard_by_split_points
@ -196,10 +311,10 @@ SELECT * FROM show_catalog;
Schema | Name | Owner
---------------------------------------------------------------------
citus_split_shard_by_split_points_negative | table_second | postgres
citus_split_shard_by_split_points_negative | table_second_102 | postgres
citus_split_shard_by_split_points_negative | table_second_103 | postgres
citus_split_shard_by_split_points_negative | table_second_2 | postgres
citus_split_shard_by_split_points_negative | table_to_split | postgres
citus_split_shard_by_split_points_negative | table_to_split_1 | postgres
citus_split_shard_by_split_points_negative | table_to_split_100 | postgres
citus_split_shard_by_split_points_negative | table_to_split_101 | postgres
(6 rows)
@ -209,14 +324,8 @@ SELECT * FROM show_catalog;
Schema | Name | Owner
---------------------------------------------------------------------
citus_split_shard_by_split_points_negative | table_second | postgres
citus_split_shard_by_split_points_negative | table_second_102 | postgres
citus_split_shard_by_split_points_negative | table_second_103 | postgres
citus_split_shard_by_split_points_negative | table_second_2 | postgres
citus_split_shard_by_split_points_negative | table_to_split | postgres
citus_split_shard_by_split_points_negative | table_to_split_1 | postgres
citus_split_shard_by_split_points_negative | table_to_split_100 | postgres
citus_split_shard_by_split_points_negative | table_to_split_101 | postgres
(8 rows)
(2 rows)
SELECT * FROM pg_publication_tables;
pubname | schemaname | tablename

View File

@ -49,14 +49,12 @@ SELECT citus_split_shard_by_split_points(
1,
ARRAY['-1073741826'],
ARRAY[:worker_2_node, :worker_2_node],
'non_blocking');
'force_logical');
-- On worker2, we want child shard 2 and dummy shard 1 --
-- on worker1, we want child shard 3 and 1 and dummy shard 2 --
\c - - - :worker_2_port
SET search_path TO citus_split_shard_by_split_points_negative;
SELECT * FROM pg_stat_subscription;
SELECT * FROM pg_subscription_rel;
SELECT * FROM show_catalog;
\c - - - :worker_1_port