/*
* LockRelationDistributionMetadata returns after getting a the lock used for a
* relation's distribution metadata, blocking if required. Only ExclusiveLock
* and ShareLock modes are supported. Any locks acquired using this method are
* released at transaction end.
*/
void
LockRelationDistributionMetadata(Oid relationId, LOCKMODE lockMode)
{
Assert(lockMode == ExclusiveLock || lockMode == ShareLock);
(void) LockRelationOid(relationId, lockMode);
}
/*
* Open the sequence and acquire AccessShareLock if needed
*
* If we haven't touched the sequence already in this transaction,
* we need to acquire AccessShareLock. We arrange for the lock to
* be owned by the top transaction, so that we don't need to do it
* more than once per xact.
*/
static Relation
open_share_lock(SeqTable seq)
{
LocalTransactionId thislxid = MyProc->lxid;
/* Get the lock if not already held in this xact */
if (seq->lxid != thislxid)
{
ResourceOwner currentOwner;
currentOwner = CurrentResourceOwner;
PG_TRY();
{
CurrentResourceOwner = TopTransactionResourceOwner;
LockRelationOid(seq->relid, AccessShareLock);
}
PG_CATCH();
{
/* Ensure CurrentResourceOwner is restored on error */
CurrentResourceOwner = currentOwner;
PG_RE_THROW();
}
PG_END_TRY();
CurrentResourceOwner = currentOwner;
/* Flag that we have a lock in the current xact */
seq->lxid = thislxid;
}
/* We now know we have AccessShareLock, and can safely open the rel */
return relation_open(seq->relid, NoLock);
}
/*
* master_drop_all_shards attempts to drop all shards for a given relation.
* Unlike master_apply_delete_command, this function can be called even
* if the table has already been dropped.
*/
Datum
master_drop_all_shards(PG_FUNCTION_ARGS)
{
Oid relationId = PG_GETARG_OID(0);
text *schemaNameText = PG_GETARG_TEXT_P(1);
text *relationNameText = PG_GETARG_TEXT_P(2);
List *shardIntervalList = NIL;
int droppedShardCount = 0;
char *schemaName = text_to_cstring(schemaNameText);
char *relationName = text_to_cstring(relationNameText);
CheckCitusVersion(ERROR);
/*
* The SQL_DROP trigger calls this function even for tables that are
* not distributed. In that case, silently ignore and return -1.
*/
if (!IsDistributedTable(relationId) || !EnableDDLPropagation)
{
PG_RETURN_INT32(-1);
}
EnsureCoordinator();
CheckTableSchemaNameForDrop(relationId, &schemaName, &relationName);
/*
* master_drop_all_shards is typically called from the DROP TABLE trigger,
* but could be called by a user directly. Make sure we have an
* AccessExlusiveLock to prevent any other commands from running on this table
* concurrently.
*/
LockRelationOid(relationId, AccessExclusiveLock);
shardIntervalList = LoadShardIntervalList(relationId);
droppedShardCount = DropShards(relationId, schemaName, relationName,
shardIntervalList);
PG_RETURN_INT32(droppedShardCount);
}
/**
* @fn Datum get_table_and_inheritors(PG_FUNCTION_ARGS)
* @brief Return array containing Oids of parent table and its children.
* Note that this function does not release relation locks.
*
* get_table_and_inheritors(table)
*
* @param table parent table.
* @retval regclass[]
*/
Datum
repack_get_table_and_inheritors(PG_FUNCTION_ARGS)
{
Oid parent = PG_GETARG_OID(0);
List *relations;
Datum *relations_array;
int relations_array_size;
ArrayType *result;
ListCell *lc;
int i;
LockRelationOid(parent, AccessShareLock);
/* Check that parent table exists */
if (!SearchSysCacheExists1(RELOID, ObjectIdGetDatum(parent)))
PG_RETURN_ARRAYTYPE_P(construct_empty_array(OIDOID));
/* Also check that children exist */
relations = find_all_inheritors(parent, AccessShareLock, NULL);
relations_array_size = list_length(relations);
if (relations_array_size == 0)
PG_RETURN_ARRAYTYPE_P(construct_empty_array(OIDOID));
relations_array = palloc(relations_array_size * sizeof(Datum));
i = 0;
foreach (lc, relations)
relations_array[i++] = ObjectIdGetDatum(lfirst_oid(lc));
result = construct_array(relations_array,
relations_array_size,
OIDOID, sizeof(Oid),
true, 'i');
pfree(relations_array);
PG_RETURN_ARRAYTYPE_P(result);
}
/*
* RecoverPreparedTransactions recovers any pending prepared
* transactions started by this node on other nodes.
*/
static int
RecoverPreparedTransactions(void)
{
List *workerList = NIL;
ListCell *workerNodeCell = NULL;
int recoveredTransactionCount = 0;
/*
* We block here if metadata transactions are ongoing, since we
* mustn't commit/abort their prepared transactions under their
* feet. We also prevent concurrent recovery.
*/
LockRelationOid(DistTransactionRelationId(), ExclusiveLock);
workerList = WorkerNodeList();
foreach(workerNodeCell, workerList)
{
WorkerNode *workerNode = (WorkerNode *) lfirst(workerNodeCell);
recoveredTransactionCount += RecoverWorkerTransactions(workerNode);
}
/*
* InitCatCachePhase2 -- external interface for CatalogCacheInitializeCache
*
* One reason to call this routine is to ensure that the relcache has
* created entries for all the catalogs and indexes referenced by catcaches.
* Therefore, provide an option to open the index as well as fixing the
* cache itself. An exception is the indexes on pg_am, which we don't use
* (cf. IndexScanOK).
*/
void
InitCatCachePhase2(CatCache *cache, bool touch_index)
{
if (cache->cc_tupdesc == NULL)
CatalogCacheInitializeCache(cache);
if (touch_index &&
cache->id != AMOID &&
cache->id != AMNAME)
{
Relation idesc;
/*
* We must lock the underlying catalog before opening the index to
* avoid deadlock, since index_open could possibly result in reading
* this same catalog, and if anyone else is exclusive-locking this
* catalog and index they'll be doing it in that order.
*/
LockRelationOid(cache->cc_reloid, AccessShareLock);
idesc = index_open(cache->cc_indexoid, AccessShareLock);
index_close(idesc, AccessShareLock);
UnlockRelationOid(cache->cc_reloid, AccessShareLock);
}
}
/*
* find_inheritance_children
*
* Returns a list containing the OIDs of all relations which
* inherit *directly* from the relation with OID 'parentrelId'.
*
* The specified lock type is acquired on each child relation (but not on the
* given rel; caller should already have locked it). If lockmode is NoLock
* then no locks are acquired, but caller must beware of race conditions
* against possible DROPs of child relations.
*/
List *
find_inheritance_children(Oid parentrelId, LOCKMODE lockmode)
{
List *list = NIL;
Relation relation;
SysScanDesc scan;
ScanKeyData key[1];
HeapTuple inheritsTuple;
Oid inhrelid;
Oid *oidarr;
int maxoids,
numoids,
i;
/*
* Can skip the scan if pg_class shows the relation has never had a
* subclass.
*/
if (!has_subclass(parentrelId))
return NIL;
/*
* Scan pg_inherits and build a working array of subclass OIDs.
*/
maxoids = 32;
oidarr = (Oid *) palloc(maxoids * sizeof(Oid));
numoids = 0;
relation = heap_open(InheritsRelationId, AccessShareLock);
ScanKeyInit(&key[0],
Anum_pg_inherits_inhparent,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(parentrelId));
scan = systable_beginscan(relation, InheritsParentIndexId, true,
NULL, 1, key);
while ((inheritsTuple = systable_getnext(scan)) != NULL)
{
inhrelid = ((Form_pg_inherits) GETSTRUCT(inheritsTuple))->inhrelid;
if (numoids >= maxoids)
{
maxoids *= 2;
oidarr = (Oid *) repalloc(oidarr, maxoids * sizeof(Oid));
}
oidarr[numoids++] = inhrelid;
}
systable_endscan(scan);
heap_close(relation, AccessShareLock);
/*
* If we found more than one child, sort them by OID. This ensures
* reasonably consistent behavior regardless of the vagaries of an
* indexscan. This is important since we need to be sure all backends
* lock children in the same order to avoid needless deadlocks.
*/
if (numoids > 1)
qsort(oidarr, numoids, sizeof(Oid), oid_cmp);
/*
* Acquire locks and build the result list.
*/
for (i = 0; i < numoids; i++)
{
inhrelid = oidarr[i];
if (lockmode != NoLock)
{
/* Get the lock to synchronize against concurrent drop */
LockRelationOid(inhrelid, lockmode);
/*
* Now that we have the lock, double-check to see if the relation
* really exists or not. If not, assume it was dropped while we
* waited to acquire lock, and ignore it.
*/
if (!SearchSysCacheExists1(RELOID, ObjectIdGetDatum(inhrelid)))
{
/* Release useless lock */
UnlockRelationOid(inhrelid, lockmode);
/* And ignore this relation */
continue;
}
}
list = lappend_oid(list, inhrelid);
}
pfree(oidarr);
return list;
}
/*
* CreateShardsWithRoundRobinPolicy creates empty shards for the given table
* based on the specified number of initial shards. The function first updates
* metadata on the coordinator node to make this shard (and its placements)
* visible. Note that the function assumes the table is hash partitioned and
* calculates the min/max hash token ranges for each shard, giving them an equal
* split of the hash space. Finally, function creates empty shard placements on
* worker nodes.
*/
void
CreateShardsWithRoundRobinPolicy(Oid distributedTableId, int32 shardCount,
int32 replicationFactor, bool useExclusiveConnections)
{
char shardStorageType = 0;
List *workerNodeList = NIL;
int32 workerNodeCount = 0;
uint32 placementAttemptCount = 0;
uint64 hashTokenIncrement = 0;
List *existingShardList = NIL;
int64 shardIndex = 0;
DistTableCacheEntry *cacheEntry = DistributedTableCacheEntry(distributedTableId);
bool colocatedShard = false;
List *insertedShardPlacements = NIL;
/* make sure table is hash partitioned */
CheckHashPartitionedTable(distributedTableId);
/*
* In contrast to append/range partitioned tables it makes more sense to
* require ownership privileges - shards for hash-partitioned tables are
* only created once, not continually during ingest as for the other
* partitioning types.
*/
EnsureTableOwner(distributedTableId);
/* we plan to add shards: get an exclusive lock on relation oid */
LockRelationOid(distributedTableId, ExclusiveLock);
/* validate that shards haven't already been created for this table */
existingShardList = LoadShardList(distributedTableId);
if (existingShardList != NIL)
{
char *tableName = get_rel_name(distributedTableId);
ereport(ERROR, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("table \"%s\" has already had shards created for it",
tableName)));
}
/* make sure that at least one shard is specified */
if (shardCount <= 0)
{
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("shard_count must be positive")));
}
/* make sure that at least one replica is specified */
if (replicationFactor <= 0)
{
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("replication_factor must be positive")));
}
/* make sure that RF=1 if the table is streaming replicated */
if (cacheEntry->replicationModel == REPLICATION_MODEL_STREAMING &&
replicationFactor > 1)
{
char *relationName = get_rel_name(cacheEntry->relationId);
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("using replication factor %d with the streaming "
"replication model is not supported",
replicationFactor),
errdetail("The table %s is marked as streaming replicated and "
"the shard replication factor of streaming replicated "
"tables must be 1.", relationName),
errhint("Use replication factor 1.")));
}
/* calculate the split of the hash space */
hashTokenIncrement = HASH_TOKEN_COUNT / shardCount;
/* don't allow concurrent node list changes that require an exclusive lock */
LockRelationOid(DistNodeRelationId(), RowShareLock);
/* load and sort the worker node list for deterministic placement */
workerNodeList = ActivePrimaryNodeList();
workerNodeList = SortList(workerNodeList, CompareWorkerNodes);
/* make sure we don't process cancel signals until all shards are created */
HOLD_INTERRUPTS();
workerNodeCount = list_length(workerNodeList);
if (replicationFactor > workerNodeCount)
{
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("replication_factor (%d) exceeds number of worker nodes "
"(%d)", replicationFactor, workerNodeCount),
errhint("Add more worker nodes or try again with a lower "
"replication factor.")));
}
/* if we have enough nodes, add an extra placement attempt for backup */
placementAttemptCount = (uint32) replicationFactor;
if (workerNodeCount > replicationFactor)
{
placementAttemptCount++;
}
/* set shard storage type according to relation type */
shardStorageType = ShardStorageType(distributedTableId);
for (shardIndex = 0; shardIndex < shardCount; shardIndex++)
{
uint32 roundRobinNodeIndex = shardIndex % workerNodeCount;
/* initialize the hash token space for this shard */
text *minHashTokenText = NULL;
text *maxHashTokenText = NULL;
int32 shardMinHashToken = INT32_MIN + (shardIndex * hashTokenIncrement);
int32 shardMaxHashToken = shardMinHashToken + (hashTokenIncrement - 1);
uint64 shardId = GetNextShardId();
List *currentInsertedShardPlacements = NIL;
/* if we are at the last shard, make sure the max token value is INT_MAX */
if (shardIndex == (shardCount - 1))
{
shardMaxHashToken = INT32_MAX;
}
/* insert the shard metadata row along with its min/max values */
minHashTokenText = IntegerToText(shardMinHashToken);
maxHashTokenText = IntegerToText(shardMaxHashToken);
/*
* Grabbing the shard metadata lock isn't technically necessary since
* we already hold an exclusive lock on the partition table, but we'll
* acquire it for the sake of completeness. As we're adding new active
* placements, the mode must be exclusive.
*/
LockShardDistributionMetadata(shardId, ExclusiveLock);
InsertShardRow(distributedTableId, shardId, shardStorageType,
minHashTokenText, maxHashTokenText);
currentInsertedShardPlacements = InsertShardPlacementRows(distributedTableId,
shardId,
workerNodeList,
roundRobinNodeIndex,
replicationFactor);
insertedShardPlacements = list_concat(insertedShardPlacements,
currentInsertedShardPlacements);
}
CreateShardsOnWorkers(distributedTableId, insertedShardPlacements,
useExclusiveConnections, colocatedShard);
if (QueryCancelPending)
{
ereport(WARNING, (errmsg("cancel requests are ignored during shard creation")));
QueryCancelPending = false;
}
RESUME_INTERRUPTS();
}
/*
* CreateColocatedShards creates shards for the target relation colocated with
* the source relation.
*/
void
CreateColocatedShards(Oid targetRelationId, Oid sourceRelationId, bool
useExclusiveConnections)
{
char targetShardStorageType = 0;
List *existingShardList = NIL;
List *sourceShardIntervalList = NIL;
ListCell *sourceShardCell = NULL;
bool colocatedShard = true;
List *insertedShardPlacements = NIL;
/* make sure that tables are hash partitioned */
CheckHashPartitionedTable(targetRelationId);
CheckHashPartitionedTable(sourceRelationId);
/*
* In contrast to append/range partitioned tables it makes more sense to
* require ownership privileges - shards for hash-partitioned tables are
* only created once, not continually during ingest as for the other
* partitioning types.
*/
EnsureTableOwner(targetRelationId);
/* we plan to add shards: get an exclusive lock on target relation oid */
LockRelationOid(targetRelationId, ExclusiveLock);
/* we don't want source table to get dropped before we colocate with it */
LockRelationOid(sourceRelationId, AccessShareLock);
/* prevent placement changes of the source relation until we colocate with them */
sourceShardIntervalList = LoadShardIntervalList(sourceRelationId);
LockShardListMetadata(sourceShardIntervalList, ShareLock);
/* validate that shards haven't already been created for this table */
existingShardList = LoadShardList(targetRelationId);
if (existingShardList != NIL)
{
char *targetRelationName = get_rel_name(targetRelationId);
ereport(ERROR, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("table \"%s\" has already had shards created for it",
targetRelationName)));
}
targetShardStorageType = ShardStorageType(targetRelationId);
foreach(sourceShardCell, sourceShardIntervalList)
{
ShardInterval *sourceShardInterval = (ShardInterval *) lfirst(sourceShardCell);
uint64 sourceShardId = sourceShardInterval->shardId;
uint64 newShardId = GetNextShardId();
ListCell *sourceShardPlacementCell = NULL;
int32 shardMinValue = DatumGetInt32(sourceShardInterval->minValue);
int32 shardMaxValue = DatumGetInt32(sourceShardInterval->maxValue);
text *shardMinValueText = IntegerToText(shardMinValue);
text *shardMaxValueText = IntegerToText(shardMaxValue);
List *sourceShardPlacementList = ShardPlacementList(sourceShardId);
InsertShardRow(targetRelationId, newShardId, targetShardStorageType,
shardMinValueText, shardMaxValueText);
foreach(sourceShardPlacementCell, sourceShardPlacementList)
{
ShardPlacement *sourcePlacement =
(ShardPlacement *) lfirst(sourceShardPlacementCell);
uint32 groupId = sourcePlacement->groupId;
const RelayFileState shardState = FILE_FINALIZED;
const uint64 shardSize = 0;
uint64 shardPlacementId = 0;
ShardPlacement *shardPlacement = NULL;
/*
* Optimistically add shard placement row the pg_dist_shard_placement, in case
* of any error it will be roll-backed.
*/
shardPlacementId = InsertShardPlacementRow(newShardId, INVALID_PLACEMENT_ID,
shardState, shardSize, groupId);
shardPlacement = LoadShardPlacement(newShardId, shardPlacementId);
insertedShardPlacements = lappend(insertedShardPlacements, shardPlacement);
}
/*
* ExecRefreshMatView -- execute a REFRESH MATERIALIZED VIEW command
*
* This refreshes the materialized view by creating a new table and swapping
* the relfilenodes of the new table and the old materialized view, so the OID
* of the original materialized view is preserved. Thus we do not lose GRANT
* nor references to this materialized view.
*
* If WITH NO DATA was specified, this is effectively like a TRUNCATE;
* otherwise it is like a TRUNCATE followed by an INSERT using the SELECT
* statement associated with the materialized view. The statement node's
* skipData field shows whether the clause was used.
*
* Indexes are rebuilt too, via REINDEX. Since we are effectively bulk-loading
* the new heap, it's better to create the indexes afterwards than to fill them
* incrementally while we load.
*
* The matview's "populated" state is changed based on whether the contents
* reflect the result set of the materialized view's query.
*/
void
ExecRefreshMatView(RefreshMatViewStmt *stmt, const char *queryString,
ParamListInfo params, char *completionTag)
{
Oid matviewOid;
Relation matviewRel;
RewriteRule *rule;
List *actions;
Query *dataQuery;
Oid tableSpace;
Oid relowner;
Oid OIDNewHeap;
DestReceiver *dest;
bool concurrent;
LOCKMODE lockmode;
Oid save_userid;
int save_sec_context;
int save_nestlevel;
/* Determine strength of lock needed. */
concurrent = stmt->concurrent;
lockmode = concurrent ? ExclusiveLock : AccessExclusiveLock;
/*
* Get a lock until end of transaction.
*/
matviewOid = RangeVarGetRelidExtended(stmt->relation,
lockmode, false, false,
RangeVarCallbackOwnsTable, NULL);
matviewRel = heap_open(matviewOid, NoLock);
/* Make sure it is a materialized view. */
if (matviewRel->rd_rel->relkind != RELKIND_MATVIEW)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("\"%s\" is not a materialized view",
RelationGetRelationName(matviewRel))));
/* Check that CONCURRENTLY is not specified if not populated. */
if (concurrent && !RelationIsPopulated(matviewRel))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("CONCURRENTLY cannot be used when the materialized view is not populated")));
/* Check that conflicting options have not been specified. */
if (concurrent && stmt->skipData)
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("CONCURRENTLY and WITH NO DATA options cannot be used together")));
/* We don't allow an oid column for a materialized view. */
Assert(!matviewRel->rd_rel->relhasoids);
/*
* Check that everything is correct for a refresh. Problems at this point
* are internal errors, so elog is sufficient.
*/
if (matviewRel->rd_rel->relhasrules == false ||
matviewRel->rd_rules->numLocks < 1)
elog(ERROR,
"materialized view \"%s\" is missing rewrite information",
RelationGetRelationName(matviewRel));
if (matviewRel->rd_rules->numLocks > 1)
elog(ERROR,
"materialized view \"%s\" has too many rules",
RelationGetRelationName(matviewRel));
rule = matviewRel->rd_rules->rules[0];
if (rule->event != CMD_SELECT || !(rule->isInstead))
elog(ERROR,
"the rule for materialized view \"%s\" is not a SELECT INSTEAD OF rule",
RelationGetRelationName(matviewRel));
actions = rule->actions;
if (list_length(actions) != 1)
elog(ERROR,
"the rule for materialized view \"%s\" is not a single action",
RelationGetRelationName(matviewRel));
/*
* The stored query was rewritten at the time of the MV definition, but
* has not been scribbled on by the planner.
*/
dataQuery = (Query *) linitial(actions);
Assert(IsA(dataQuery, Query));
/*
* Check for active uses of the relation in the current transaction, such
* as open scans.
*
* NB: We count on this to protect us against problems with refreshing the
* data using HEAP_INSERT_FROZEN.
*/
CheckTableNotInUse(matviewRel, "REFRESH MATERIALIZED VIEW");
/*
* Tentatively mark the matview as populated or not (this will roll back
* if we fail later).
*/
SetMatViewPopulatedState(matviewRel, !stmt->skipData);
relowner = matviewRel->rd_rel->relowner;
/*
* Switch to the owner's userid, so that any functions are run as that
* user. Also arrange to make GUC variable changes local to this command.
* Don't lock it down too tight to create a temporary table just yet. We
* will switch modes when we are about to execute user code.
*/
GetUserIdAndSecContext(&save_userid, &save_sec_context);
SetUserIdAndSecContext(relowner,
save_sec_context | SECURITY_LOCAL_USERID_CHANGE);
save_nestlevel = NewGUCNestLevel();
/* Concurrent refresh builds new data in temp tablespace, and does diff. */
if (concurrent)
tableSpace = GetDefaultTablespace(RELPERSISTENCE_TEMP);
else
tableSpace = matviewRel->rd_rel->reltablespace;
/*
* Create the transient table that will receive the regenerated data. Lock
* it against access by any other process until commit (by which time it
* will be gone).
*/
OIDNewHeap = make_new_heap(matviewOid, tableSpace, concurrent,
ExclusiveLock);
LockRelationOid(OIDNewHeap, AccessExclusiveLock);
dest = CreateTransientRelDestReceiver(OIDNewHeap);
/*
* Now lock down security-restricted operations.
*/
SetUserIdAndSecContext(relowner,
save_sec_context | SECURITY_RESTRICTED_OPERATION);
/* Generate the data, if wanted. */
if (!stmt->skipData)
refresh_matview_datafill(dest, dataQuery, queryString);
heap_close(matviewRel, NoLock);
/* Make the matview match the newly generated data. */
if (concurrent)
{
int old_depth = matview_maintenance_depth;
PG_TRY();
{
refresh_by_match_merge(matviewOid, OIDNewHeap, relowner,
save_sec_context);
}
PG_CATCH();
{
matview_maintenance_depth = old_depth;
PG_RE_THROW();
}
PG_END_TRY();
Assert(matview_maintenance_depth == old_depth);
}
else
refresh_by_heap_swap(matviewOid, OIDNewHeap);
/* Roll back any GUC changes */
AtEOXact_GUC(false, save_nestlevel);
/* Restore userid and security context */
SetUserIdAndSecContext(save_userid, save_sec_context);
}
void
InitResQueues(void)
{
HeapTuple tuple;
int numQueues = 0;
bool queuesok = true;
cqContext *pcqCtx;
cqContext cqc;
Assert(ResScheduler);
/*
* Need a resource owner to keep the heapam code happy.
*/
Assert(CurrentResourceOwner == NULL);
ResourceOwner owner = ResourceOwnerCreate(NULL, "InitQueues");
CurrentResourceOwner = owner;
/**
* The resqueue shared mem initialization must be serialized. Only the first session
* should do the init.
* Serialization is done the ResQueueLock LW_EXCLUSIVE. However, we must obtain all DB
* lock before obtaining LWlock.
* So, we must have obtained ResQueueRelationId and ResQueueCapabilityRelationId lock
* first.
*/
Relation relResqueue = heap_open(ResQueueRelationId, AccessShareLock);
LockRelationOid(ResQueueCapabilityRelationId, RowExclusiveLock);
LWLockAcquire(ResQueueLock, LW_EXCLUSIVE);
if (ResScheduler->num_queues > 0)
{
/* Hash table has already been loaded */
LWLockRelease(ResQueueLock);
UnlockRelationOid(ResQueueCapabilityRelationId, RowExclusiveLock);
heap_close(relResqueue, AccessShareLock);
CurrentResourceOwner = NULL;
ResourceOwnerDelete(owner);
return;
}
/* XXX XXX: should this be rowexclusive ? */
pcqCtx = caql_beginscan(
caql_indexOK(
caql_addrel(cqclr(&cqc), relResqueue),
false),
cql("SELECT * FROM pg_resqueue ", NULL));
while (HeapTupleIsValid(tuple = caql_getnext(pcqCtx)))
{
Form_pg_resqueue queueform;
Oid queueid;
bool overcommit;
float4 ignorelimit;
Cost thresholds[NUM_RES_LIMIT_TYPES];
char *queuename;
numQueues++;
queueform = (Form_pg_resqueue) GETSTRUCT(tuple);
queueid = HeapTupleGetOid(tuple);
queuename = NameStr(queueform->rsqname);
thresholds[RES_COUNT_LIMIT] = queueform->rsqcountlimit;
thresholds[RES_COST_LIMIT] = queueform->rsqcostlimit;
thresholds[RES_MEMORY_LIMIT] = ResourceQueueGetMemoryLimit(queueid);
overcommit = queueform->rsqovercommit;
ignorelimit = queueform->rsqignorecostlimit;
queuesok = ResCreateQueue(queueid, thresholds, overcommit, ignorelimit);
if (!queuesok)
{
/** Break out of loop. Close relations, relinquish LWLock and then error out */
break;
}
}
caql_endscan(pcqCtx);
LWLockRelease(ResQueueLock);
UnlockRelationOid(ResQueueCapabilityRelationId, RowExclusiveLock);
heap_close(relResqueue, AccessShareLock);
if (!queuesok)
ereport(PANIC,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("insufficient resource queues available"),
errhint("Increase max_resource_queues to %d.", numQueues)));
elog(LOG,"initialized %d resource queues", numQueues);
CurrentResourceOwner = NULL;
ResourceOwnerDelete(owner);
return;
}
/*
* Alter table space move
*
* Allows a user to move all of their objects in a given tablespace in the
* current database to another tablespace. Only objects which the user is
* considered to be an owner of are moved and the user must have CREATE rights
* on the new tablespace. These checks should mean that ALTER TABLE will never
* fail due to permissions, but note that permissions will also be checked at
* that level. Objects can be ALL, TABLES, INDEXES, or MATERIALIZED VIEWS.
*
* All to-be-moved objects are locked first. If NOWAIT is specified and the
* lock can't be acquired then we ereport(ERROR).
*/
Oid
AlterTableSpaceMove(AlterTableSpaceMoveStmt *stmt)
{
List *relations = NIL;
ListCell *l;
ScanKeyData key[1];
Relation rel;
HeapScanDesc scan;
HeapTuple tuple;
Oid orig_tablespaceoid;
Oid new_tablespaceoid;
List *role_oids = roleNamesToIds(stmt->roles);
/* Ensure we were not asked to move something we can't */
if (!stmt->move_all && stmt->objtype != OBJECT_TABLE &&
stmt->objtype != OBJECT_INDEX && stmt->objtype != OBJECT_MATVIEW)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("only tables, indexes, and materialized views exist in tablespaces")));
/* Get the orig and new tablespace OIDs */
orig_tablespaceoid = get_tablespace_oid(stmt->orig_tablespacename, false);
new_tablespaceoid = get_tablespace_oid(stmt->new_tablespacename, false);
/* Can't move shared relations in to or out of pg_global */
/* This is also checked by ATExecSetTableSpace, but nice to stop earlier */
if (orig_tablespaceoid == GLOBALTABLESPACE_OID ||
new_tablespaceoid == GLOBALTABLESPACE_OID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("cannot move relations in to or out of pg_global tablespace")));
/*
* Must have CREATE rights on the new tablespace, unless it is the
* database default tablespace (which all users implicitly have CREATE
* rights on).
*/
if (OidIsValid(new_tablespaceoid) && new_tablespaceoid != MyDatabaseTableSpace)
{
AclResult aclresult;
aclresult = pg_tablespace_aclcheck(new_tablespaceoid, GetUserId(),
ACL_CREATE);
if (aclresult != ACLCHECK_OK)
aclcheck_error(aclresult, ACL_KIND_TABLESPACE,
get_tablespace_name(new_tablespaceoid));
}
/*
* Now that the checks are done, check if we should set either to
* InvalidOid because it is our database's default tablespace.
*/
if (orig_tablespaceoid == MyDatabaseTableSpace)
orig_tablespaceoid = InvalidOid;
if (new_tablespaceoid == MyDatabaseTableSpace)
new_tablespaceoid = InvalidOid;
/* no-op */
if (orig_tablespaceoid == new_tablespaceoid)
return new_tablespaceoid;
/*
* Walk the list of objects in the tablespace and move them. This will
* only find objects in our database, of course.
*/
ScanKeyInit(&key[0],
Anum_pg_class_reltablespace,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(orig_tablespaceoid));
rel = heap_open(RelationRelationId, AccessShareLock);
scan = heap_beginscan_catalog(rel, 1, key);
while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
Oid relOid = HeapTupleGetOid(tuple);
Form_pg_class relForm;
relForm = (Form_pg_class) GETSTRUCT(tuple);
/*
* Do not move objects in pg_catalog as part of this, if an admin
* really wishes to do so, they can issue the individual ALTER
* commands directly.
*
* Also, explicitly avoid any shared tables, temp tables, or TOAST
* (TOAST will be moved with the main table).
*/
if (IsSystemNamespace(relForm->relnamespace) || relForm->relisshared ||
isAnyTempNamespace(relForm->relnamespace) ||
relForm->relnamespace == PG_TOAST_NAMESPACE)
continue;
/* Only consider objects which live in tablespaces */
if (relForm->relkind != RELKIND_RELATION &&
relForm->relkind != RELKIND_INDEX &&
relForm->relkind != RELKIND_MATVIEW)
continue;
/* Check if we were asked to only move a certain type of object */
if (!stmt->move_all &&
((stmt->objtype == OBJECT_TABLE &&
relForm->relkind != RELKIND_RELATION) ||
(stmt->objtype == OBJECT_INDEX &&
relForm->relkind != RELKIND_INDEX) ||
(stmt->objtype == OBJECT_MATVIEW &&
relForm->relkind != RELKIND_MATVIEW)))
continue;
/* Check if we are only moving objects owned by certain roles */
if (role_oids != NIL && !list_member_oid(role_oids, relForm->relowner))
continue;
/*
* Handle permissions-checking here since we are locking the tables
* and also to avoid doing a bunch of work only to fail part-way.
* Note that permissions will also be checked by AlterTableInternal().
*
* Caller must be considered an owner on the table to move it.
*/
if (!pg_class_ownercheck(relOid, GetUserId()))
aclcheck_error(ACLCHECK_NOT_OWNER, ACL_KIND_CLASS,
NameStr(relForm->relname));
if (stmt->nowait &&
!ConditionalLockRelationOid(relOid, AccessExclusiveLock))
ereport(ERROR,
(errcode(ERRCODE_OBJECT_IN_USE),
errmsg("aborting due to \"%s\".\"%s\" --- lock not available",
get_namespace_name(relForm->relnamespace),
NameStr(relForm->relname))));
else
LockRelationOid(relOid, AccessExclusiveLock);
/* Add to our list of objects to move */
relations = lappend_oid(relations, relOid);
}
heap_endscan(scan);
heap_close(rel, AccessShareLock);
if (relations == NIL)
ereport(NOTICE,
(errcode(ERRCODE_NO_DATA_FOUND),
errmsg("no matching relations in tablespace \"%s\" found",
orig_tablespaceoid == InvalidOid ? "(database default)" :
get_tablespace_name(orig_tablespaceoid))));
/* Everything is locked, loop through and move all of the relations. */
foreach(l, relations)
{
List *cmds = NIL;
AlterTableCmd *cmd = makeNode(AlterTableCmd);
cmd->subtype = AT_SetTableSpace;
cmd->name = stmt->new_tablespacename;
cmds = lappend(cmds, cmd);
AlterTableInternal(lfirst_oid(l), cmds, false);
}
