/*
* Returns a count of the number of non-template databases from the catalog.
*/
int get_database_count(void) {
int retval, processed;
StringInfoData buf;
SPITupleTable *coltuptable;
int database_count = 0;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
initStringInfo(&buf);
appendStringInfo(&buf, "SELECT count(*) FROM pg_database WHERE datname NOT IN ('template0', 'template1') AND datallowconn IS TRUE;");
retval = SPI_execute(buf.data, false, 0);
if (retval != SPI_OK_SELECT) {
elog(FATAL, "Database information collection failed");
// FAIL RETURN 1
}
processed = SPI_processed;
if (processed > 0) {
coltuptable = SPI_tuptable;
database_count = atoi(SPI_getvalue(coltuptable->vals[0], coltuptable->tupdesc, 1));
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
return database_count;
}
/*
* Main worker routine. Accepts dsm_handle as an argument
*/
static void
bg_worker_main(Datum main_arg)
{
PartitionArgs *args;
dsm_handle handle = DatumGetInt32(main_arg);
/* Create resource owner */
CurrentResourceOwner = ResourceOwnerCreate(NULL, "CreatePartitionsWorker");
/* Attach to dynamic shared memory */
if (!handle)
{
ereport(WARNING,
(errmsg("pg_pathman worker: invalid dsm_handle")));
}
segment = dsm_attach(handle);
args = dsm_segment_address(segment);
/* Establish connection and start transaction */
BackgroundWorkerInitializeConnectionByOid(args->dbid, InvalidOid);
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
/* Create partitions */
args->result = create_partitions(args->relid, PATHMAN_GET_DATUM(args->value, args->by_val), args->value_type, &args->crashed);
/* Cleanup */
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
dsm_detach(segment);
}
void
SetEStateSnapshot(EState *estate, ResourceOwner owner)
{
estate->es_snapshot = GetTransactionSnapshot();
estate->es_snapshot->active_count++;
estate->es_snapshot->copied = true;
PushActiveSnapshot(estate->es_snapshot);
}
/*
* Ensure that the environment is sane.
* This involves checking the Postgresql version, and if in network mode
* also establishing a connection to a receiver.
*/
int ensure_valid_environment(void) {
StringInfoData buf;
int retval;
char* pgversion;
SPITupleTable *coltuptable;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
/* Ensure compatible version */
pgstat_report_activity(STATE_RUNNING, "verifying compatible postgres version");
initStringInfo(&buf);
appendStringInfo(&buf,
"select version();"
);
retval = SPI_execute(buf.data, false, 0);
if (retval != SPI_OK_SELECT) {
elog(FATAL, "Unable to query postgres version %d", retval);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
return 1;
}
coltuptable = SPI_tuptable;
pgversion = SPI_getvalue(coltuptable->vals[0], coltuptable->tupdesc, 1);
if(strstr(pgversion, "PostgreSQL 9.3") == NULL) {
elog(FATAL, "Unsupported Postgresql version");
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
return 1;
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
/*
* Attempt to establish a connection if the output mode is network.
*/
if (strcmp(output_mode, "network") == 0) {
retval = establish_connection();
if (retval == 2) {
elog(LOG, "Error : Failed to connect to antenna please check domain is available from host.");
}
}
//TODO verify logging directory is accessible when csv mode.
elog(LOG, "Pgsampler Initialized");
return 0;
}
/*
* Initialize workspace for a worker process: create the schema if it doesn't
* already exist.
*/
static void
initialize_worker_spi(worktable *table)
{
int ret;
int ntup;
bool isnull;
StringInfoData buf;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "initializing spi_worker schema");
/* XXX could we use CREATE SCHEMA IF NOT EXISTS? */
initStringInfo(&buf);
appendStringInfo(&buf, "select count(*) from pg_namespace where nspname = '%s'",
table->schema);
ret = SPI_execute(buf.data, true, 0);
if (ret != SPI_OK_SELECT)
elog(FATAL, "SPI_execute failed: error code %d", ret);
if (SPI_processed != 1)
elog(FATAL, "not a singleton result");
ntup = DatumGetInt64(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (isnull)
elog(FATAL, "null result");
if (ntup == 0)
{
resetStringInfo(&buf);
appendStringInfo(&buf,
"CREATE SCHEMA \"%s\" "
"CREATE TABLE \"%s\" ("
" type text CHECK (type IN ('total', 'delta')), "
" value integer)"
"CREATE UNIQUE INDEX \"%s_unique_total\" ON \"%s\" (type) "
"WHERE type = 'total'",
table->schema, table->name, table->name, table->name);
/* set statement start time */
SetCurrentStatementStartTimestamp();
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_UTILITY)
elog(FATAL, "failed to create my schema");
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
static void
set_snapshot(EState *estate, ResourceOwner owner)
{
estate->es_snapshot = GetTransactionSnapshot();
estate->es_snapshot->active_count++;
estate->es_snapshot->copied = true;
RegisterSnapshotOnOwner(estate->es_snapshot, owner);
PushActiveSnapshot(estate->es_snapshot);
}
/*
* Handle INSERT message.
*/
static void
apply_handle_insert(StringInfo s)
{
LogicalRepRelMapEntry *rel;
LogicalRepTupleData newtup;
LogicalRepRelId relid;
EState *estate;
TupleTableSlot *remoteslot;
MemoryContext oldctx;
ensure_transaction();
relid = logicalrep_read_insert(s, &newtup);
rel = logicalrep_rel_open(relid, RowExclusiveLock);
if (!should_apply_changes_for_rel(rel))
{
/*
* The relation can't become interesting in the middle of the
* transaction so it's safe to unlock it.
*/
logicalrep_rel_close(rel, RowExclusiveLock);
return;
}
/* Initialize the executor state. */
estate = create_estate_for_relation(rel);
remoteslot = ExecInitExtraTupleSlot(estate,
RelationGetDescr(rel->localrel));
/* Input functions may need an active snapshot, so get one */
PushActiveSnapshot(GetTransactionSnapshot());
/* Process and store remote tuple in the slot */
oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
slot_store_cstrings(remoteslot, rel, newtup.values);
slot_fill_defaults(rel, estate, remoteslot);
MemoryContextSwitchTo(oldctx);
ExecOpenIndices(estate->es_result_relation_info, false);
/* Do the insert. */
ExecSimpleRelationInsert(estate, remoteslot);
/* Cleanup. */
ExecCloseIndices(estate->es_result_relation_info);
PopActiveSnapshot();
/* Handle queued AFTER triggers. */
AfterTriggerEndQuery(estate);
ExecResetTupleTable(estate->es_tupleTable, false);
FreeExecutorState(estate);
logicalrep_rel_close(rel, NoLock);
CommandCounterIncrement();
}
static void
execute_pg_settings_logger(config_log_objects *objects) {
int ret;
bool isnull;
StringInfoData buf;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "executing configuration logger function");
initStringInfo(&buf);
appendStringInfo(
&buf,
"SELECT %s.%s()",
config_log_schema,
objects->function_name
);
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_SELECT)
{
elog(FATAL, "SPI_execute failed: error code %d", ret);
}
if (SPI_processed != 1)
{
elog(FATAL, "not a singleton result");
}
log_info("pg_settings_logger() executed");
if(DatumGetBool(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull)))
{
log_info("Configuration changes recorded");
}
else
{
log_info("No configuration changes detected");
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
void worker_main(Datum arg)
{
int ret;
StringInfoData buf;
uint32 segment = UInt32GetDatum(arg);
/* Setup signal handlers */
pqsignal(SIGHUP, worker_sighup);
pqsignal(SIGTERM, worker_sigterm);
/* Allow signals */
BackgroundWorkerUnblockSignals();
initialize_worker(segment);
/* Connect to the database */
BackgroundWorkerInitializeConnection(job->datname, job->rolname);
elog(LOG, "%s initialized running job id %d", MyBgworkerEntry->bgw_name, job->job_id);
pgstat_report_appname(MyBgworkerEntry->bgw_name);
/* Initialize the query text */
initStringInfo(&buf);
appendStringInfo(&buf,
"SELECT * FROM %s.%s(%d, NULL)",
job_run_function.schema,
job_run_function.name,
job->job_id);
/* Initialize the SPI subsystem */
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
SetCurrentStatementStartTimestamp();
/* And run the query */
ret = SPI_execute(buf.data, true, 0);
if (ret < 0)
elog(FATAL, "errors while executing %s", buf.data);
/* Commmit the transaction */
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
proc_exit(0);
}
/*
* This function will set a string in shared memory which is the name of the database to connect to
* the next time the background worker restarts. Because a bgworker can only connect to one database
* at a time, and some catalogs and stats are scoped to the current database, the bg worker
* periodically restarts to collect latest stats from another database.
*
*/
int set_next_db_target(void) {
int retval, processed;
StringInfoData buf;
SPITupleTable *coltuptable;
char* next_db_target;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
/* get sorted list of databases, find one after target_db*/
initStringInfo(&buf);
appendStringInfo(&buf,
"SELECT datname FROM pg_database WHERE datname NOT IN ('template0', 'template1') AND datallowconn IS TRUE AND datname > '%s' ORDER BY datname ASC LIMIT 1;", target_db
);
retval = SPI_execute(buf.data, false, 0);
if (retval != SPI_OK_SELECT) {
elog(FATAL, "Database information collection failed");
// FAIL RETURN 1
}
processed = SPI_processed;
if(processed == 0) {
//No matching records so pick first database.
resetStringInfo(&buf);
appendStringInfoString(&buf,
"SELECT datname FROM pg_database WHERE datname NOT IN ('template0', 'template1') AND datallowconn IS TRUE ORDER BY datname ASC LIMIT 1;"
);
retval = SPI_execute(buf.data, false, 0);
if (retval != SPI_OK_SELECT) {
elog(FATAL, "Database information collection failed");
// FAIL RETURN 1
}
}
coltuptable = SPI_tuptable;
next_db_target = SPI_getvalue(coltuptable->vals[0], coltuptable->tupdesc, 1);
// elog(LOG, "NEXTDB TARGET: %s", next_db_target); //print next target db
strcpy(pgsampler_state->next_db, next_db_target);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
return 0;
}
/*
* PushUpdatedSnapshot
* As above, except we set the snapshot's CID to the current CID.
*/
void
PushUpdatedSnapshot(Snapshot snapshot)
{
Snapshot newsnap;
/*
* We cannot risk modifying a snapshot that's possibly already used
* elsewhere, so make a new copy to scribble on.
*/
newsnap = CopySnapshot(snapshot);
newsnap->curcid = GetCurrentCommandId(false);
PushActiveSnapshot(newsnap);
}
/*
* Wait until the relation synchronization state is set in the catalog to the
* expected one.
*
* Used when transitioning from CATCHUP state to SYNCDONE.
*
* Returns false if the synchronization worker has disappeared or the table state
* has been reset.
*/
static bool
wait_for_relation_state_change(Oid relid, char expected_state)
{
char state;
for (;;)
{
LogicalRepWorker *worker;
XLogRecPtr statelsn;
CHECK_FOR_INTERRUPTS();
/* XXX use cache invalidation here to improve performance? */
PushActiveSnapshot(GetLatestSnapshot());
state = GetSubscriptionRelState(MyLogicalRepWorker->subid,
relid, &statelsn, true);
PopActiveSnapshot();
if (state == SUBREL_STATE_UNKNOWN)
return false;
if (state == expected_state)
return true;
/* Check if the sync worker is still running and bail if not. */
LWLockAcquire(LogicalRepWorkerLock, LW_SHARED);
/* Check if the opposite worker is still running and bail if not. */
worker = logicalrep_worker_find(MyLogicalRepWorker->subid,
am_tablesync_worker() ? InvalidOid : relid,
false);
LWLockRelease(LogicalRepWorkerLock);
if (!worker)
return false;
(void) WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
1000L, WAIT_EVENT_LOGICAL_SYNC_STATE_CHANGE);
ResetLatch(MyLatch);
}
return false;
}
/*
* CopyIntoStream
*
* COPY events to a stream from an input source
*/
void
CopyIntoStream(Relation rel, TupleDesc desc, HeapTuple *tuples, int ntuples)
{
bool snap = ActiveSnapshotSet();
ResultRelInfo rinfo;
StreamInsertState *sis;
MemSet(&rinfo, 0, sizeof(ResultRelInfo));
rinfo.ri_RangeTableIndex = 1; /* dummy */
rinfo.ri_TrigDesc = NULL;
rinfo.ri_RelationDesc = rel;
if (snap)
PopActiveSnapshot();
BeginStreamModify(NULL, &rinfo, list_make1(desc), 0, 0);
sis = (StreamInsertState *) rinfo.ri_FdwState;
Assert(sis);
if (sis->queries)
{
TupleTableSlot *slot = MakeSingleTupleTableSlot(RelationGetDescr(rel));
int i;
for (i = 0; i < ntuples; i++)
{
ExecStoreTuple(tuples[i], slot, InvalidBuffer, false);
ExecStreamInsert(NULL, &rinfo, slot, NULL);
ExecClearTuple(slot);
}
ExecDropSingleTupleTableSlot(slot);
Assert(sis->ntups == ntuples);
pgstat_increment_cq_write(ntuples, sis->nbytes);
}
EndStreamModify(NULL, &rinfo);
if (snap)
PushActiveSnapshot(GetTransactionSnapshot());
}
/*
* Handle DELETE message.
*
* TODO: FDW support
*/
static void
apply_handle_delete(StringInfo s)
{
LogicalRepRelMapEntry *rel;
LogicalRepTupleData oldtup;
LogicalRepRelId relid;
Oid idxoid;
EState *estate;
EPQState epqstate;
TupleTableSlot *remoteslot;
TupleTableSlot *localslot;
bool found;
MemoryContext oldctx;
ensure_transaction();
relid = logicalrep_read_delete(s, &oldtup);
rel = logicalrep_rel_open(relid, RowExclusiveLock);
if (!should_apply_changes_for_rel(rel))
{
/*
* The relation can't become interesting in the middle of the
* transaction so it's safe to unlock it.
*/
logicalrep_rel_close(rel, RowExclusiveLock);
return;
}
/* Check if we can do the delete. */
check_relation_updatable(rel);
/* Initialize the executor state. */
estate = create_estate_for_relation(rel);
remoteslot = ExecInitExtraTupleSlot(estate,
RelationGetDescr(rel->localrel));
localslot = ExecInitExtraTupleSlot(estate,
RelationGetDescr(rel->localrel));
EvalPlanQualInit(&epqstate, estate, NULL, NIL, -1);
PushActiveSnapshot(GetTransactionSnapshot());
ExecOpenIndices(estate->es_result_relation_info, false);
/* Find the tuple using the replica identity index. */
oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
slot_store_cstrings(remoteslot, rel, oldtup.values);
MemoryContextSwitchTo(oldctx);
/*
* Try to find tuple using either replica identity index, primary key or
* if needed, sequential scan.
*/
idxoid = GetRelationIdentityOrPK(rel->localrel);
Assert(OidIsValid(idxoid) ||
(rel->remoterel.replident == REPLICA_IDENTITY_FULL));
if (OidIsValid(idxoid))
found = RelationFindReplTupleByIndex(rel->localrel, idxoid,
LockTupleExclusive,
remoteslot, localslot);
else
found = RelationFindReplTupleSeq(rel->localrel, LockTupleExclusive,
remoteslot, localslot);
/* If found delete it. */
if (found)
{
EvalPlanQualSetSlot(&epqstate, localslot);
/* Do the actual delete. */
ExecSimpleRelationDelete(estate, &epqstate, localslot);
}
else
{
/* The tuple to be deleted could not be found. */
ereport(DEBUG1,
(errmsg("logical replication could not find row for delete "
"in replication target relation \"%s\"",
RelationGetRelationName(rel->localrel))));
}
/* Cleanup. */
ExecCloseIndices(estate->es_result_relation_info);
PopActiveSnapshot();
/* Handle queued AFTER triggers. */
AfterTriggerEndQuery(estate);
EvalPlanQualEnd(&epqstate);
ExecResetTupleTable(estate->es_tupleTable, false);
FreeExecutorState(estate);
logicalrep_rel_close(rel, NoLock);
CommandCounterIncrement();
}
/*
* Handle UPDATE message.
*
* TODO: FDW support
*/
static void
apply_handle_update(StringInfo s)
{
LogicalRepRelMapEntry *rel;
LogicalRepRelId relid;
Oid idxoid;
EState *estate;
EPQState epqstate;
LogicalRepTupleData oldtup;
LogicalRepTupleData newtup;
bool has_oldtup;
TupleTableSlot *localslot;
TupleTableSlot *remoteslot;
bool found;
MemoryContext oldctx;
ensure_transaction();
relid = logicalrep_read_update(s, &has_oldtup, &oldtup,
&newtup);
rel = logicalrep_rel_open(relid, RowExclusiveLock);
if (!should_apply_changes_for_rel(rel))
{
/*
* The relation can't become interesting in the middle of the
* transaction so it's safe to unlock it.
*/
logicalrep_rel_close(rel, RowExclusiveLock);
return;
}
/* Check if we can do the update. */
check_relation_updatable(rel);
/* Initialize the executor state. */
estate = create_estate_for_relation(rel);
remoteslot = ExecInitExtraTupleSlot(estate,
RelationGetDescr(rel->localrel));
localslot = ExecInitExtraTupleSlot(estate,
RelationGetDescr(rel->localrel));
EvalPlanQualInit(&epqstate, estate, NULL, NIL, -1);
PushActiveSnapshot(GetTransactionSnapshot());
ExecOpenIndices(estate->es_result_relation_info, false);
/* Build the search tuple. */
oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
slot_store_cstrings(remoteslot, rel,
has_oldtup ? oldtup.values : newtup.values);
MemoryContextSwitchTo(oldctx);
/*
* Try to find tuple using either replica identity index, primary key or
* if needed, sequential scan.
*/
idxoid = GetRelationIdentityOrPK(rel->localrel);
Assert(OidIsValid(idxoid) ||
(rel->remoterel.replident == REPLICA_IDENTITY_FULL && has_oldtup));
if (OidIsValid(idxoid))
found = RelationFindReplTupleByIndex(rel->localrel, idxoid,
LockTupleExclusive,
remoteslot, localslot);
else
found = RelationFindReplTupleSeq(rel->localrel, LockTupleExclusive,
remoteslot, localslot);
ExecClearTuple(remoteslot);
/*
* Tuple found.
*
* Note this will fail if there are other conflicting unique indexes.
*/
if (found)
{
/* Process and store remote tuple in the slot */
oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
ExecStoreTuple(localslot->tts_tuple, remoteslot, InvalidBuffer, false);
slot_modify_cstrings(remoteslot, rel, newtup.values, newtup.changed);
MemoryContextSwitchTo(oldctx);
EvalPlanQualSetSlot(&epqstate, remoteslot);
/* Do the actual update. */
ExecSimpleRelationUpdate(estate, &epqstate, localslot, remoteslot);
}
else
{
/*
* The tuple to be updated could not be found.
*
* TODO what to do here, change the log level to LOG perhaps?
*/
elog(DEBUG1,
"logical replication did not find row for update "
"in replication target relation \"%s\"",
RelationGetRelationName(rel->localrel));
}
/* Cleanup. */
ExecCloseIndices(estate->es_result_relation_info);
PopActiveSnapshot();
/* Handle queued AFTER triggers. */
AfterTriggerEndQuery(estate);
EvalPlanQualEnd(&epqstate);
ExecResetTupleTable(estate->es_tupleTable, false);
FreeExecutorState(estate);
logicalrep_rel_close(rel, NoLock);
CommandCounterIncrement();
}
/*
* PushCopiedSnapshot
* As above, except forcibly copy the presented snapshot.
*
* This should be used when the ActiveSnapshot has to be modifiable, for
* example if the caller intends to call UpdateActiveSnapshotCommandId.
* The new snapshot will be released when popped from the stack.
*/
void
PushCopiedSnapshot(Snapshot snapshot)
{
PushActiveSnapshot(CopySnapshot(snapshot));
}
void
worker_test_main(Datum main_arg)
{
dsm_segment *seg;
volatile test_shm_mq_header *hdr;
PGPROC *registrant;
pqsignal(SIGHUP, handle_sighup);
pqsignal(SIGTERM, handle_sigterm);
BackgroundWorkerUnblockSignals();
printf("worker_test_main: %d\n", DatumGetInt32(main_arg));
CurrentResourceOwner = ResourceOwnerCreate(NULL, "worker test");
seg = dsm_attach(DatumGetInt32(main_arg));
if (seg == NULL)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("unable to map dynamic shared memory segment")));
hdr = dsm_segment_address(seg);
/* 開始 */
SpinLockAcquire(&hdr->mutex);
hdr->workers_ready++;
hdr->workers_attached++;
SpinLockRelease(&hdr->mutex);
registrant = BackendPidGetProc(MyBgworkerEntry->bgw_notify_pid);
if (registrant == NULL)
{
elog(DEBUG1, "registrant backend has exited prematurely");
proc_exit(1);
}
SetLatch(®istrant->procLatch);
/* Do the work */
BackgroundWorkerInitializeConnection(hdr->dbname, NULL);
printf("DSM: %p\n", dsm_segment_address);
#if 0
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "initializing spi_worker schema");
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
#endif
dsm_detach(seg);
proc_exit(0);
}
/*
* PersistHoldablePortal
*
* Prepare the specified Portal for access outside of the current
* transaction. When this function returns, all future accesses to the
* portal must be done via the Tuplestore (not by invoking the
* executor).
*/
void
PersistHoldablePortal(Portal portal)
{
QueryDesc *queryDesc = PortalGetQueryDesc(portal);
Portal saveActivePortal;
ResourceOwner saveResourceOwner;
MemoryContext savePortalContext;
MemoryContext oldcxt;
/*
* If we're preserving a holdable portal, we had better be inside the
* transaction that originally created it.
*/
Assert(portal->createSubid != InvalidSubTransactionId);
Assert(queryDesc != NULL);
/*
* Caller must have created the tuplestore already ... but not a snapshot.
*/
Assert(portal->holdContext != NULL);
Assert(portal->holdStore != NULL);
Assert(portal->holdSnapshot == NULL);
/*
* Before closing down the executor, we must copy the tupdesc into
* long-term memory, since it was created in executor memory.
*/
oldcxt = MemoryContextSwitchTo(portal->holdContext);
portal->tupDesc = CreateTupleDescCopy(portal->tupDesc);
MemoryContextSwitchTo(oldcxt);
/*
* Check for improper portal use, and mark portal active.
*/
MarkPortalActive(portal);
/*
* Set up global portal context pointers.
*/
saveActivePortal = ActivePortal;
saveResourceOwner = CurrentResourceOwner;
savePortalContext = PortalContext;
PG_TRY();
{
ActivePortal = portal;
if (portal->resowner)
CurrentResourceOwner = portal->resowner;
PortalContext = PortalGetHeapMemory(portal);
MemoryContextSwitchTo(PortalContext);
PushActiveSnapshot(queryDesc->snapshot);
/*
* Rewind the executor: we need to store the entire result set in the
* tuplestore, so that subsequent backward FETCHs can be processed.
*/
ExecutorRewind(queryDesc);
/*
* Change the destination to output to the tuplestore. Note we tell
* the tuplestore receiver to detoast all data passed through it; this
* makes it safe to not keep a snapshot associated with the data.
*/
queryDesc->dest = CreateDestReceiver(DestTuplestore);
SetTuplestoreDestReceiverParams(queryDesc->dest,
portal->holdStore,
portal->holdContext,
true);
/* Fetch the result set into the tuplestore */
ExecutorRun(queryDesc, ForwardScanDirection, 0L);
(*queryDesc->dest->rDestroy) (queryDesc->dest);
queryDesc->dest = NULL;
/*
* Now shut down the inner executor.
*/
portal->queryDesc = NULL; /* prevent double shutdown */
ExecutorFinish(queryDesc);
ExecutorEnd(queryDesc);
FreeQueryDesc(queryDesc);
/*
* Set the position in the result set.
*/
MemoryContextSwitchTo(portal->holdContext);
if (portal->atEnd)
{
/*
* Just force the tuplestore forward to its end. The size of the
* skip request here is arbitrary.
*/
while (tuplestore_skiptuples(portal->holdStore, 1000000, true))
/* continue */ ;
}
else
{
tuplestore_rescan(portal->holdStore);
if (!tuplestore_skiptuples(portal->holdStore,
portal->portalPos,
true))
elog(ERROR, "unexpected end of tuple stream");
}
}
PG_CATCH();
{
/* Uncaught error while executing portal: mark it dead */
MarkPortalFailed(portal);
/* Restore global vars and propagate error */
ActivePortal = saveActivePortal;
CurrentResourceOwner = saveResourceOwner;
PortalContext = savePortalContext;
PG_RE_THROW();
}
PG_END_TRY();
MemoryContextSwitchTo(oldcxt);
/* Mark portal not active */
portal->status = PORTAL_READY;
ActivePortal = saveActivePortal;
CurrentResourceOwner = saveResourceOwner;
PortalContext = savePortalContext;
PopActiveSnapshot();
/*
* We can now release any subsidiary memory of the portal's heap context;
* we'll never use it again. The executor already dropped its context,
* but this will clean up anything that glommed onto the portal's heap via
* PortalContext.
*/
MemoryContextDeleteChildren(PortalGetHeapMemory(portal));
}
static config_log_objects *
initialize_objects(void)
{
config_log_objects *objects;
int ret;
int ntup;
bool isnull;
StringInfoData buf;
objects = palloc(sizeof(config_log_objects));
objects->table_name = pstrdup("pg_settings_log");
objects->function_name = pstrdup("pg_settings_logger");
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "Verifying config log objects");
initStringInfo(&buf);
appendStringInfo(
&buf,
"SELECT COUNT(*)\
FROM information_schema.tables\
WHERE table_schema='%s'\
AND table_name ='%s'\
AND table_type='BASE TABLE'",
config_log_schema,
objects->table_name
);
ret = SPI_execute(buf.data, true, 0);
if (ret != SPI_OK_SELECT)
{
ereport(FATAL,
(errmsg("SPI_execute failed: SPI error code %d", ret)
));
}
/* This should never happen */
if (SPI_processed != 1)
{
elog(FATAL, "not a singleton result");
}
ntup = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
/* This should never happen */
if (isnull)
{
elog(FATAL, "null result");
}
if (ntup == 0)
{
ereport(FATAL,
(
errmsg("Expected config log table '%s.%s' not found", config_log_schema,
objects->table_name),
errhint("Ensure superuser search_path includes the schema used by config_log; "
"check config_log.* GUC settings")
));
}
/* check function pg_settings_logger() exists */
resetStringInfo(&buf);
appendStringInfo(
&buf,
"SELECT COUNT(*) FROM pg_catalog.pg_proc p \
INNER JOIN pg_catalog.pg_namespace n ON n.oid = p.pronamespace \
WHERE p.proname='%s' \
AND n.nspname='%s' \
AND p.pronargs = 0",
objects->function_name,
config_log_schema
);
ret = SPI_execute(buf.data, true, 0);
if (ret != SPI_OK_SELECT)
{
ereport(FATAL,
(errmsg("SPI_execute failed: SPI error code %d", ret)));
}
if (SPI_processed != 1)
{
elog(FATAL, "not a singleton result");
}
ntup = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (isnull)
{
elog(FATAL, "null result");
}
if (ntup == 0)
{
ereport(FATAL,
(
errmsg("Expected config log function '%s.%s' not found", config_log_schema,
objects->function_name),
errhint("Ensure superuser search_path includes the schema used by config_log; "
"check config_log.* GUC settings")
));
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
log_info("initialized, database objects validated");
/* execute pg_settings_logger() here to catch any settings which have changed after server restart */
execute_pg_settings_logger(objects);
return objects;
}
/*
* Search the relation 'rel' for tuple using the sequential scan.
*
* If a matching tuple is found, lock it with lockmode, fill the slot with its
* contents, and return true. Return false otherwise.
*
* Note that this stops on the first matching tuple.
*
* This can obviously be quite slow on tables that have more than few rows.
*/
bool
RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode,
TupleTableSlot *searchslot, TupleTableSlot *outslot)
{
HeapTuple scantuple;
HeapScanDesc scan;
SnapshotData snap;
TransactionId xwait;
bool found;
TupleDesc desc = RelationGetDescr(rel);
Assert(equalTupleDescs(desc, outslot->tts_tupleDescriptor));
/* Start an index scan. */
InitDirtySnapshot(snap);
scan = heap_beginscan(rel, &snap, 0, NULL);
retry:
found = false;
heap_rescan(scan, NULL);
/* Try to find the tuple */
while ((scantuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
if (!tuple_equals_slot(desc, scantuple, searchslot))
continue;
found = true;
ExecStoreTuple(scantuple, outslot, InvalidBuffer, false);
ExecMaterializeSlot(outslot);
xwait = TransactionIdIsValid(snap.xmin) ?
snap.xmin : snap.xmax;
/*
* If the tuple is locked, wait for locking transaction to finish and
* retry.
*/
if (TransactionIdIsValid(xwait))
{
XactLockTableWait(xwait, NULL, NULL, XLTW_None);
goto retry;
}
}
/* Found tuple, try to lock it in the lockmode. */
if (found)
{
Buffer buf;
HeapUpdateFailureData hufd;
HTSU_Result res;
HeapTupleData locktup;
ItemPointerCopy(&outslot->tts_tuple->t_self, &locktup.t_self);
PushActiveSnapshot(GetLatestSnapshot());
res = heap_lock_tuple(rel, &locktup, GetCurrentCommandId(false),
lockmode,
LockWaitBlock,
false /* don't follow updates */ ,
&buf, &hufd);
/* the tuple slot already has the buffer pinned */
ReleaseBuffer(buf);
PopActiveSnapshot();
switch (res)
{
case HeapTupleMayBeUpdated:
break;
case HeapTupleUpdated:
/* XXX: Improve handling here */
ereport(LOG,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("concurrent update, retrying")));
goto retry;
case HeapTupleInvisible:
elog(ERROR, "attempted to lock invisible tuple");
default:
elog(ERROR, "unexpected heap_lock_tuple status: %u", res);
break;
}
}
heap_endscan(scan);
return found;
}
/*
* Search the relation 'rel' for tuple using the index.
*
* If a matching tuple is found, lock it with lockmode, fill the slot with its
* contents, and return true. Return false otherwise.
*/
bool
RelationFindReplTupleByIndex(Relation rel, Oid idxoid,
LockTupleMode lockmode,
TupleTableSlot *searchslot,
TupleTableSlot *outslot)
{
HeapTuple scantuple;
ScanKeyData skey[INDEX_MAX_KEYS];
IndexScanDesc scan;
SnapshotData snap;
TransactionId xwait;
Relation idxrel;
bool found;
/* Open the index. */
idxrel = index_open(idxoid, RowExclusiveLock);
/* Start an index scan. */
InitDirtySnapshot(snap);
scan = index_beginscan(rel, idxrel, &snap,
RelationGetNumberOfAttributes(idxrel),
0);
/* Build scan key. */
build_replindex_scan_key(skey, rel, idxrel, searchslot);
retry:
found = false;
index_rescan(scan, skey, RelationGetNumberOfAttributes(idxrel), NULL, 0);
/* Try to find the tuple */
if ((scantuple = index_getnext(scan, ForwardScanDirection)) != NULL)
{
found = true;
ExecStoreTuple(scantuple, outslot, InvalidBuffer, false);
ExecMaterializeSlot(outslot);
xwait = TransactionIdIsValid(snap.xmin) ?
snap.xmin : snap.xmax;
/*
* If the tuple is locked, wait for locking transaction to finish and
* retry.
*/
if (TransactionIdIsValid(xwait))
{
XactLockTableWait(xwait, NULL, NULL, XLTW_None);
goto retry;
}
}
/* Found tuple, try to lock it in the lockmode. */
if (found)
{
Buffer buf;
HeapUpdateFailureData hufd;
HTSU_Result res;
HeapTupleData locktup;
ItemPointerCopy(&outslot->tts_tuple->t_self, &locktup.t_self);
PushActiveSnapshot(GetLatestSnapshot());
res = heap_lock_tuple(rel, &locktup, GetCurrentCommandId(false),
lockmode,
LockWaitBlock,
false /* don't follow updates */ ,
&buf, &hufd);
/* the tuple slot already has the buffer pinned */
ReleaseBuffer(buf);
PopActiveSnapshot();
switch (res)
{
case HeapTupleMayBeUpdated:
break;
case HeapTupleUpdated:
/* XXX: Improve handling here */
ereport(LOG,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("concurrent update, retrying")));
goto retry;
case HeapTupleInvisible:
elog(ERROR, "attempted to lock invisible tuple");
default:
elog(ERROR, "unexpected heap_lock_tuple status: %u", res);
break;
}
}
index_endscan(scan);
/* Don't release lock until commit. */
index_close(idxrel, NoLock);
return found;
}
/*
* Primary entry point for VACUUM and ANALYZE commands.
*
* options is a bitmask of VacuumOption flags, indicating what to do.
*
* relid, if not InvalidOid, indicate the relation to process; otherwise,
* the RangeVar is used. (The latter must always be passed, because it's
* used for error messages.)
*
* params contains a set of parameters that can be used to customize the
* behavior.
*
* va_cols is a list of columns to analyze, or NIL to process them all.
*
* bstrategy is normally given as NULL, but in autovacuum it can be passed
* in to use the same buffer strategy object across multiple vacuum() calls.
*
* isTopLevel should be passed down from ProcessUtility.
*
* It is the caller's responsibility that all parameters are allocated in a
* memory context that will not disappear at transaction commit.
*/
void
vacuum(int options, RangeVar *relation, Oid relid, VacuumParams *params,
List *va_cols, BufferAccessStrategy bstrategy, bool isTopLevel)
{
const char *stmttype;
volatile bool in_outer_xact,
use_own_xacts;
List *relations;
static bool in_vacuum = false;
Assert(params != NULL);
stmttype = (options & VACOPT_VACUUM) ? "VACUUM" : "ANALYZE";
/*
* We cannot run VACUUM inside a user transaction block; if we were inside
* a transaction, then our commit- and start-transaction-command calls
* would not have the intended effect! There are numerous other subtle
* dependencies on this, too.
*
* ANALYZE (without VACUUM) can run either way.
*/
if (options & VACOPT_VACUUM)
{
PreventTransactionChain(isTopLevel, stmttype);
in_outer_xact = false;
}
else
in_outer_xact = IsInTransactionChain(isTopLevel);
/*
* Due to static variables vac_context, anl_context and vac_strategy,
* vacuum() is not reentrant. This matters when VACUUM FULL or ANALYZE
* calls a hostile index expression that itself calls ANALYZE.
*/
if (in_vacuum)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("%s cannot be executed from VACUUM or ANALYZE",
stmttype)));
/*
* Sanity check DISABLE_PAGE_SKIPPING option.
*/
if ((options & VACOPT_FULL) != 0 &&
(options & VACOPT_DISABLE_PAGE_SKIPPING) != 0)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("VACUUM option DISABLE_PAGE_SKIPPING cannot be used with FULL")));
/*
* Send info about dead objects to the statistics collector, unless we are
* in autovacuum --- autovacuum.c does this for itself.
*/
if ((options & VACOPT_VACUUM) && !IsAutoVacuumWorkerProcess())
pgstat_vacuum_stat();
/*
* Create special memory context for cross-transaction storage.
*
* Since it is a child of PortalContext, it will go away eventually even
* if we suffer an error; there's no need for special abort cleanup logic.
*/
vac_context = AllocSetContextCreate(PortalContext,
"Vacuum",
ALLOCSET_DEFAULT_SIZES);
/*
* If caller didn't give us a buffer strategy object, make one in the
* cross-transaction memory context.
*/
if (bstrategy == NULL)
{
MemoryContext old_context = MemoryContextSwitchTo(vac_context);
bstrategy = GetAccessStrategy(BAS_VACUUM);
MemoryContextSwitchTo(old_context);
}
vac_strategy = bstrategy;
/*
* Build list of relations to process, unless caller gave us one. (If we
* build one, we put it in vac_context for safekeeping.)
*/
relations = get_rel_oids(relid, relation);
/*
* Decide whether we need to start/commit our own transactions.
*
* For VACUUM (with or without ANALYZE): always do so, so that we can
* release locks as soon as possible. (We could possibly use the outer
* transaction for a one-table VACUUM, but handling TOAST tables would be
* problematic.)
*
* For ANALYZE (no VACUUM): if inside a transaction block, we cannot
* start/commit our own transactions. Also, there's no need to do so if
* only processing one relation. For multiple relations when not within a
* transaction block, and also in an autovacuum worker, use own
* transactions so we can release locks sooner.
*/
if (options & VACOPT_VACUUM)
use_own_xacts = true;
else
{
Assert(options & VACOPT_ANALYZE);
if (IsAutoVacuumWorkerProcess())
use_own_xacts = true;
else if (in_outer_xact)
use_own_xacts = false;
else if (list_length(relations) > 1)
use_own_xacts = true;
else
use_own_xacts = false;
}
/*
* vacuum_rel expects to be entered with no transaction active; it will
* start and commit its own transaction. But we are called by an SQL
* command, and so we are executing inside a transaction already. We
* commit the transaction started in PostgresMain() here, and start
* another one before exiting to match the commit waiting for us back in
* PostgresMain().
*/
if (use_own_xacts)
{
Assert(!in_outer_xact);
/* ActiveSnapshot is not set by autovacuum */
if (ActiveSnapshotSet())
PopActiveSnapshot();
/* matches the StartTransaction in PostgresMain() */
CommitTransactionCommand();
}
/* Turn vacuum cost accounting on or off */
PG_TRY();
{
ListCell *cur;
in_vacuum = true;
VacuumCostActive = (VacuumCostDelay > 0);
VacuumCostBalance = 0;
VacuumPageHit = 0;
VacuumPageMiss = 0;
VacuumPageDirty = 0;
/*
* Loop to process each selected relation.
*/
foreach(cur, relations)
{
Oid relid = lfirst_oid(cur);
if (options & VACOPT_VACUUM)
{
if (!vacuum_rel(relid, relation, options, params))
continue;
}
if (options & VACOPT_ANALYZE)
{
/*
* If using separate xacts, start one for analyze. Otherwise,
* we can use the outer transaction.
*/
if (use_own_xacts)
{
StartTransactionCommand();
/* functions in indexes may want a snapshot set */
PushActiveSnapshot(GetTransactionSnapshot());
}
analyze_rel(relid, relation, options, params,
va_cols, in_outer_xact, vac_strategy);
if (use_own_xacts)
{
PopActiveSnapshot();
CommitTransactionCommand();
}
}
}
}
void
worker_spi_main(Datum main_arg)
{
int index = DatumGetInt32(main_arg);
worktable *table;
StringInfoData buf;
char name[20];
table = palloc(sizeof(worktable));
sprintf(name, "schema%d", index);
table->schema = pstrdup(name);
table->name = pstrdup("counted");
/* Establish signal handlers before unblocking signals. */
pqsignal(SIGHUP, worker_spi_sighup);
pqsignal(SIGTERM, worker_spi_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection("postgres", NULL);
elog(LOG, "%s initialized with %s.%s",
MyBgworkerEntry->bgw_name, table->schema, table->name);
initialize_worker_spi(table);
/*
* Quote identifiers passed to us. Note that this must be done after
* initialize_worker_spi, because that routine assumes the names are not
* quoted.
*
* Note some memory might be leaked here.
*/
table->schema = quote_identifier(table->schema);
table->name = quote_identifier(table->name);
initStringInfo(&buf);
appendStringInfo(&buf,
"WITH deleted AS (DELETE "
"FROM %s.%s "
"WHERE type = 'delta' RETURNING value), "
"total AS (SELECT coalesce(sum(value), 0) as sum "
"FROM deleted) "
"UPDATE %s.%s "
"SET value = %s.value + total.sum "
"FROM total WHERE type = 'total' "
"RETURNING %s.value",
table->schema, table->name,
table->schema, table->name,
table->name,
table->name);
/*
* Main loop: do this until the SIGTERM handler tells us to terminate
*/
while (!got_sigterm)
{
int ret;
int rc;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
worker_spi_naptime * 1000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/*
* In case of a SIGHUP, just reload the configuration.
*/
if (got_sighup)
{
got_sighup = false;
ProcessConfigFile(PGC_SIGHUP);
}
/*
* Start a transaction on which we can run queries. Note that each
* StartTransactionCommand() call should be preceded by a
* SetCurrentStatementStartTimestamp() call, which sets both the time
* for the statement we're about the run, and also the transaction
* start time. Also, each other query sent to SPI should probably be
* preceded by SetCurrentStatementStartTimestamp(), so that statement
* start time is always up to date.
*
* The SPI_connect() call lets us run queries through the SPI manager,
* and the PushActiveSnapshot() call creates an "active" snapshot
* which is necessary for queries to have MVCC data to work on.
*
* The pgstat_report_activity() call makes our activity visible
* through the pgstat views.
*/
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
/* We can now execute queries via SPI */
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_UPDATE_RETURNING)
elog(FATAL, "cannot select from table %s.%s: error code %d",
table->schema, table->name, ret);
if (SPI_processed > 0)
{
bool isnull;
int32 val;
val = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (!isnull)
elog(LOG, "%s: count in %s.%s is now %d",
MyBgworkerEntry->bgw_name,
table->schema, table->name, val);
}
/*
* And finish our transaction.
*/
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
proc_exit(1);
}
/*
* worker logic
*/
void
wed_worker_main(Datum main_arg)
{
StringInfoData buf;
/* Establish signal handlers before unblocking signals. */
pqsignal(SIGHUP, wed_worker_sighup);
pqsignal(SIGTERM, wed_worker_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection(wed_worker_db_name, NULL);
elog(LOG, "%s initialized in: %s",
MyBgworkerEntry->bgw_name, wed_worker_db_name);
initStringInfo(&buf);
appendStringInfo(&buf, "SELECT trcheck()");
/*
* Main loop: do this until the SIGTERM handler tells us to terminate
*/
while (!got_sigterm)
{
int ret;
int rc;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
wed_worker_naptime * 1000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/*
* In case of a SIGHUP, just reload the configuration.
*/
if (got_sighup)
{
got_sighup = false;
ProcessConfigFile(PGC_SIGHUP);
}
/*
* Start a transaction on which we can run queries. Note that each
* StartTransactionCommand() call should be preceded by a
* SetCurrentStatementStartTimestamp() call, which sets both the time
* for the statement we're about the run, and also the transaction
* start time. Also, each other query sent to SPI should probably be
* preceded by SetCurrentStatementStartTimestamp(), so that statement
* start time is always up to date.
*
* The SPI_connect() call lets us run queries through the SPI manager,
* and the PushActiveSnapshot() call creates an "active" snapshot
* which is necessary for queries to have MVCC data to work on.
*
* The pgstat_report_activity() call makes our activity visible
* through the pgstat views.
*/
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
/* We can now execute queries via SPI */
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_SELECT)
elog(FATAL, "stored procedure trcheck() not found: error code %d", ret);
elog(LOG, "%s : trcheck() done !", MyBgworkerEntry->bgw_name);
/*
* And finish our transaction.
*/
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
proc_exit(1);
}
/*
* PersistHoldablePortal
*
* Prepare the specified Portal for access outside of the current
* transaction. When this function returns, all future accesses to the
* portal must be done via the Tuplestore (not by invoking the
* executor).
*/
void
PersistHoldablePortal(Portal portal)
{
QueryDesc *queryDesc = PortalGetQueryDesc(portal);
Portal saveActivePortal;
ResourceOwner saveResourceOwner;
MemoryContext savePortalContext;
MemoryContext oldcxt;
/*
* If we're preserving a holdable portal, we had better be inside the
* transaction that originally created it.
*/
Assert(portal->createSubid != InvalidSubTransactionId);
Assert(queryDesc != NULL);
/*
* Caller must have created the tuplestore already.
*/
Assert(portal->holdContext != NULL);
Assert(portal->holdStore != NULL);
/*
* Before closing down the executor, we must copy the tupdesc into
* long-term memory, since it was created in executor memory.
*/
oldcxt = MemoryContextSwitchTo(portal->holdContext);
portal->tupDesc = CreateTupleDescCopy(portal->tupDesc);
MemoryContextSwitchTo(oldcxt);
/*
* Check for improper portal use, and mark portal active.
*/
if (portal->status != PORTAL_READY)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("portal \"%s\" cannot be run", portal->name)));
portal->status = PORTAL_ACTIVE;
/*
* Set up global portal context pointers.
*/
saveActivePortal = ActivePortal;
saveResourceOwner = CurrentResourceOwner;
savePortalContext = PortalContext;
PG_TRY();
{
ActivePortal = portal;
CurrentResourceOwner = portal->resowner;
PortalContext = PortalGetHeapMemory(portal);
MemoryContextSwitchTo(PortalContext);
PushActiveSnapshot(queryDesc->snapshot);
/*
* Rewind the executor: we need to store the entire result set in the
* tuplestore, so that subsequent backward FETCHs can be processed.
*/
ExecutorRewind(queryDesc);
/*
* Change the destination to output to the tuplestore. Note we tell
* the tuplestore receiver to detoast all data passed through it.
*/
queryDesc->dest = CreateDestReceiver(DestTuplestore);
SetTuplestoreDestReceiverParams(queryDesc->dest,
portal->holdStore,
portal->holdContext,
true);
/* Fetch the result set into the tuplestore */
ExecutorRun(queryDesc, ForwardScanDirection, 0L);
(*queryDesc->dest->rDestroy) (queryDesc->dest);
queryDesc->dest = NULL;
/*
* Now shut down the inner executor.
*/
portal->queryDesc = NULL; /* prevent double shutdown */
/* we do not need AfterTriggerEndQuery() here */
ExecutorEnd(queryDesc);
FreeQueryDesc(queryDesc);
/*
* Set the position in the result set: ideally, this could be
* implemented by just skipping straight to the tuple # that we need
* to be at, but the tuplestore API doesn't support that. So we start
* at the beginning of the tuplestore and iterate through it until we
* reach where we need to be. FIXME someday? (Fortunately, the
* typical case is that we're supposed to be at or near the start of
* the result set, so this isn't as bad as it sounds.)
*/
MemoryContextSwitchTo(portal->holdContext);
if (portal->atEnd)
{
/* we can handle this case even if posOverflow */
while (tuplestore_advance(portal->holdStore, true))
/* continue */ ;
}
else
{
long store_pos;
if (portal->posOverflow) /* oops, cannot trust portalPos */
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("could not reposition held cursor")));
tuplestore_rescan(portal->holdStore);
for (store_pos = 0; store_pos < portal->portalPos; store_pos++)
{
if (!tuplestore_advance(portal->holdStore, true))
elog(ERROR, "unexpected end of tuple stream");
}
}
}
PG_CATCH();
{
/* Uncaught error while executing portal: mark it dead */
portal->status = PORTAL_FAILED;
/* Restore global vars and propagate error */
ActivePortal = saveActivePortal;
CurrentResourceOwner = saveResourceOwner;
PortalContext = savePortalContext;
PG_RE_THROW();
}
PG_END_TRY();
MemoryContextSwitchTo(oldcxt);
/* Mark portal not active */
portal->status = PORTAL_READY;
ActivePortal = saveActivePortal;
CurrentResourceOwner = saveResourceOwner;
PortalContext = savePortalContext;
PopActiveSnapshot();
/*
* We can now release any subsidiary memory of the portal's heap context;
* we'll never use it again. The executor already dropped its context,
* but this will clean up anything that glommed onto the portal's heap via
* PortalContext.
*/
MemoryContextDeleteChildren(PortalGetHeapMemory(portal));
}
/*
* Primary entry point for VACUUM and ANALYZE commands.
*
* relid is normally InvalidOid; if it is not, then it provides the relation
* OID to be processed, and vacstmt->relation is ignored. (The non-invalid
* case is currently only used by autovacuum.)
*
* do_toast is passed as FALSE by autovacuum, because it processes TOAST
* tables separately.
*
* for_wraparound is used by autovacuum to let us know when it's forcing
* a vacuum for wraparound, which should not be auto-cancelled.
*
* bstrategy is normally given as NULL, but in autovacuum it can be passed
* in to use the same buffer strategy object across multiple vacuum() calls.
*
* isTopLevel should be passed down from ProcessUtility.
*
* It is the caller's responsibility that vacstmt and bstrategy
* (if given) be allocated in a memory context that won't disappear
* at transaction commit.
*/
void
vacuum(VacuumStmt *vacstmt, Oid relid, bool do_toast,
BufferAccessStrategy bstrategy, bool for_wraparound, bool isTopLevel)
{
const char *stmttype;
volatile bool all_rels,
in_outer_xact,
use_own_xacts;
List *relations;
/* sanity checks on options */
Assert(vacstmt->options & (VACOPT_VACUUM | VACOPT_ANALYZE));
Assert((vacstmt->options & VACOPT_VACUUM) ||
!(vacstmt->options & (VACOPT_FULL | VACOPT_FREEZE)));
Assert((vacstmt->options & VACOPT_ANALYZE) || vacstmt->va_cols == NIL);
stmttype = (vacstmt->options & VACOPT_VACUUM) ? "VACUUM" : "ANALYZE";
/*
* We cannot run VACUUM inside a user transaction block; if we were inside
* a transaction, then our commit- and start-transaction-command calls
* would not have the intended effect! There are numerous other subtle
* dependencies on this, too.
*
* ANALYZE (without VACUUM) can run either way.
*/
if (vacstmt->options & VACOPT_VACUUM)
{
PreventTransactionChain(isTopLevel, stmttype);
in_outer_xact = false;
}
else
in_outer_xact = IsInTransactionChain(isTopLevel);
/*
* Send info about dead objects to the statistics collector, unless we are
* in autovacuum --- autovacuum.c does this for itself.
*/
if ((vacstmt->options & VACOPT_VACUUM) && !IsAutoVacuumWorkerProcess())
pgstat_vacuum_stat();
/*
* Create special memory context for cross-transaction storage.
*
* Since it is a child of PortalContext, it will go away eventually even
* if we suffer an error; there's no need for special abort cleanup logic.
*/
vac_context = AllocSetContextCreate(PortalContext,
"Vacuum",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
/*
* If caller didn't give us a buffer strategy object, make one in the
* cross-transaction memory context.
*/
if (bstrategy == NULL)
{
MemoryContext old_context = MemoryContextSwitchTo(vac_context);
bstrategy = GetAccessStrategy(BAS_VACUUM);
MemoryContextSwitchTo(old_context);
}
vac_strategy = bstrategy;
/* Remember whether we are processing everything in the DB */
all_rels = (!OidIsValid(relid) && vacstmt->relation == NULL);
/*
* Build list of relations to process, unless caller gave us one. (If we
* build one, we put it in vac_context for safekeeping.)
*/
relations = get_rel_oids(relid, vacstmt->relation);
/*
* Decide whether we need to start/commit our own transactions.
*
* For VACUUM (with or without ANALYZE): always do so, so that we can
* release locks as soon as possible. (We could possibly use the outer
* transaction for a one-table VACUUM, but handling TOAST tables would be
* problematic.)
*
* For ANALYZE (no VACUUM): if inside a transaction block, we cannot
* start/commit our own transactions. Also, there's no need to do so if
* only processing one relation. For multiple relations when not within a
* transaction block, and also in an autovacuum worker, use own
* transactions so we can release locks sooner.
*/
if (vacstmt->options & VACOPT_VACUUM)
use_own_xacts = true;
else
{
Assert(vacstmt->options & VACOPT_ANALYZE);
if (IsAutoVacuumWorkerProcess())
use_own_xacts = true;
else if (in_outer_xact)
use_own_xacts = false;
else if (list_length(relations) > 1)
use_own_xacts = true;
else
use_own_xacts = false;
}
/*
* vacuum_rel expects to be entered with no transaction active; it will
* start and commit its own transaction. But we are called by an SQL
* command, and so we are executing inside a transaction already. We
* commit the transaction started in PostgresMain() here, and start
* another one before exiting to match the commit waiting for us back in
* PostgresMain().
*/
if (use_own_xacts)
{
/* ActiveSnapshot is not set by autovacuum */
if (ActiveSnapshotSet())
PopActiveSnapshot();
/* matches the StartTransaction in PostgresMain() */
CommitTransactionCommand();
}
/* Turn vacuum cost accounting on or off */
PG_TRY();
{
ListCell *cur;
VacuumCostActive = (VacuumCostDelay > 0);
VacuumCostBalance = 0;
/*
* Loop to process each selected relation.
*/
foreach(cur, relations)
{
Oid relid = lfirst_oid(cur);
bool scanned_all = false;
if (vacstmt->options & VACOPT_VACUUM)
vacuum_rel(relid, vacstmt, do_toast, for_wraparound,
&scanned_all);
if (vacstmt->options & VACOPT_ANALYZE)
{
/*
* If using separate xacts, start one for analyze. Otherwise,
* we can use the outer transaction.
*/
if (use_own_xacts)
{
StartTransactionCommand();
/* functions in indexes may want a snapshot set */
PushActiveSnapshot(GetTransactionSnapshot());
}
analyze_rel(relid, vacstmt, vac_strategy, !scanned_all);
if (use_own_xacts)
{
PopActiveSnapshot();
CommitTransactionCommand();
}
}
}
}
static void
worker_spi_main(Datum main_arg)
{
/* Register functions for SIGTERM/SIGHUP management */
pqsignal(SIGHUP, worker_spi_sighup);
pqsignal(SIGTERM, worker_spi_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection("postgres", NULL);
while (!got_sigterm)
{
int ret;
int rc;
StringInfoData buf;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
1000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
initStringInfo(&buf);
/* Build the query string */
appendStringInfo(&buf,
"SELECT count(*) FROM pg_class;");
ret = SPI_execute(buf.data, true, 0);
/* Some error messages in case of incorrect handling */
if (ret != SPI_OK_SELECT)
elog(FATAL, "SPI_execute failed: error code %d", ret);
if (SPI_processed > 0)
{
int32 count;
bool isnull;
count = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
elog(LOG, "Currently %d relations in database",
count);
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
}
proc_exit(0);
}
static void powa_main(Datum main_arg)
{
char *q1 = "SELECT powa_take_snapshot()";
static char *q2 = "SET application_name = 'POWA collector'";
instr_time begin;
instr_time end;
long time_to_wait;
die_on_too_small_frequency();
/*
Set up signal handlers, then unblock signalsl
*/
pqsignal(SIGHUP, powa_sighup);
pqsignal(SIGTERM, powa_sigterm);
BackgroundWorkerUnblockSignals();
/*
We only connect when powa_frequency >0. If not, powa has been deactivated
*/
if (powa_frequency < 0)
{
elog(LOG, "POWA is deactivated (powa.frequency = %i), exiting",
powa_frequency);
exit(1);
}
// We got here: it means powa_frequency > 0. Let's connect
/*
Connect to POWA database
*/
BackgroundWorkerInitializeConnection(powa_database, NULL);
elog(LOG, "POWA connected to %s", powa_database);
StartTransactionCommand();
SetCurrentStatementStartTimestamp();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
SPI_execute(q2, false, 0);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
/*
let's store the current time. It will be used to
calculate a quite stable interval between each measure
*/
while (!got_sigterm)
{
/*
We can get here with a new value of powa_frequency
because of a reload. Let's suicide to disconnect
if this value is <0
*/
if (powa_frequency < 0)
{
elog(LOG, "POWA exits to disconnect from the database now");
exit(1);
}
INSTR_TIME_SET_CURRENT(begin);
ResetLatch(&MyProc->procLatch);
StartTransactionCommand();
SetCurrentStatementStartTimestamp();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
SPI_execute(q1, false, 0);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
INSTR_TIME_SET_CURRENT(end);
INSTR_TIME_SUBTRACT(end, begin);
/*
Wait powa.frequency, compensate for work time of last snapshot
*/
/*
If we got off schedule (because of a compact or delete,
just do another operation right now
*/
time_to_wait = powa_frequency - INSTR_TIME_GET_MILLISEC(end);
if (time_to_wait > 0)
{
WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
time_to_wait);
}
}
proc_exit(0);
}
/*
* CopyIntoStream
*
* COPY events to a stream from an input source
*/
void
CopyIntoStream(Relation stream, TupleDesc desc, HeapTuple *tuples, int ntuples)
{
int i;
InsertBatchAck *ack = NULL;
InsertBatch *batch = NULL;
Size size = 0;
bool snap = ActiveSnapshotSet();
Bitmapset *all_targets = GetStreamReaders(RelationGetRelid(stream));
Bitmapset *adhoc = GetAdhocContinuousViewIds();
Bitmapset *targets = bms_difference(all_targets, adhoc);
dsm_cqueue *cq = NULL;
bytea *packed_desc;
if (snap)
PopActiveSnapshot();
packed_desc = PackTupleDesc(desc);
if (!bms_is_empty(targets))
{
if (synchronous_stream_insert)
{
batch = InsertBatchCreate();
ack = palloc0(sizeof(InsertBatchAck));
ack->batch_id = batch->id;
ack->batch = batch;
}
cq = GetWorkerQueue();
}
for (i=0; i<ntuples; i++)
{
StreamTupleState *sts;
HeapTuple tup = tuples[i];
int len;
sts = StreamTupleStateCreate(tup, desc, packed_desc, targets, ack, &len);
if (cq)
{
dsm_cqueue_push_nolock(cq, sts, len);
size += len;
}
}
pfree(packed_desc);
if (cq)
dsm_cqueue_unlock(cq);
stream_stat_report(RelationGetRelid(stream), ntuples, 1, size);
if (batch)
{
pfree(ack);
InsertBatchWaitAndRemove(batch, ntuples);
}
if (snap)
PushActiveSnapshot(GetTransactionSnapshot());
bms_free(all_targets);
bms_free(adhoc);
bms_free(targets);
}
static void
kill_idle_main(Datum main_arg)
{
StringInfoData buf;
/* Register functions for SIGTERM/SIGHUP management */
pqsignal(SIGHUP, kill_idle_sighup);
pqsignal(SIGTERM, kill_idle_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to a database */
BackgroundWorkerInitializeConnection("postgres", NULL);
/* Build query for process */
initStringInfo(&buf);
kill_idle_build_query(&buf);
while (!got_sigterm)
{
int rc, ret, i;
/* Wait necessary amount of time */
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
kill_max_idle_time * 1000L,
PG_WAIT_EXTENSION);
ResetLatch(&MyProc->procLatch);
/* Emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/* Process signals */
if (got_sighup)
{
int old_interval;
/* Save old value of kill interval */
old_interval = kill_max_idle_time;
/* Process config file */
ProcessConfigFile(PGC_SIGHUP);
got_sighup = false;
ereport(LOG, (errmsg("bgworker kill_idle signal: processed SIGHUP")));
/* Rebuild query if necessary */
if (old_interval != kill_max_idle_time)
{
resetStringInfo(&buf);
initStringInfo(&buf);
kill_idle_build_query(&buf);
}
}
if (got_sigterm)
{
/* Simply exit */
ereport(LOG, (errmsg("bgworker kill_idle signal: processed SIGTERM")));
proc_exit(0);
}
/* Process idle connection kill */
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
/* Statement start time */
SetCurrentStatementStartTimestamp();
/* Execute query */
ret = SPI_execute(buf.data, false, 0);
/* Some error handling */
if (ret != SPI_OK_SELECT)
elog(FATAL, "Error when trying to kill idle connections");
/* Do some processing and log stuff disconnected */
for (i = 0; i < SPI_processed; i++)
{
int32 pidValue;
bool isnull;
char *datname = NULL;
char *usename = NULL;
char *client_addr = NULL;
/* Fetch values */
pidValue = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[i],
SPI_tuptable->tupdesc,
1, &isnull));
usename = DatumGetCString(SPI_getbinval(SPI_tuptable->vals[i],
SPI_tuptable->tupdesc,
3, &isnull));
datname = DatumGetCString(SPI_getbinval(SPI_tuptable->vals[i],
SPI_tuptable->tupdesc,
4, &isnull));
client_addr = DatumGetCString(SPI_getbinval(SPI_tuptable->vals[i],
SPI_tuptable->tupdesc,
5, &isnull));
/* Log what has been disconnected */
elog(LOG, "Disconnected idle connection: PID %d %s/%s/%s",
pidValue, datname ? datname : "none",
usename ? usename : "none",
client_addr ? client_addr : "none");
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
/* No problems, so clean exit */
proc_exit(0);
}