/*
* Create a new logical slot, with invalid LSN and xid, directly. This does not
* use the snapshot builder or logical decoding machinery. It's only intended
* for creating a slot on a replica that mirrors the state of a slot on an
* upstream master.
*
* Note that this is test harness code. You shouldn't expose slot internals
* to SQL like this for any real world usage. See the README.
*/
Datum
test_slot_timelines_create_logical_slot(PG_FUNCTION_ARGS)
{
char *slotname = text_to_cstring(PG_GETARG_TEXT_P(0));
char *plugin = text_to_cstring(PG_GETARG_TEXT_P(1));
CheckSlotRequirements();
ReplicationSlotCreate(slotname, true, RS_PERSISTENT);
/* register the plugin name with the slot */
StrNCpy(NameStr(MyReplicationSlot->data.plugin), plugin, NAMEDATALEN);
/*
* Initialize persistent state to placeholders to be set by
* test_slot_timelines_advance_logical_slot .
*/
MyReplicationSlot->data.xmin = InvalidTransactionId;
MyReplicationSlot->data.catalog_xmin = InvalidTransactionId;
MyReplicationSlot->data.restart_lsn = InvalidXLogRecPtr;
MyReplicationSlot->data.confirmed_flush = InvalidXLogRecPtr;
clear_slot_transient_state();
ReplicationSlotRelease();
PG_RETURN_VOID();
}
/*
* Set the state of a slot.
*
* This doesn't maintain the non-persistent state at all,
* but since the slot isn't in use that's OK.
*
* There's intentionally no check to prevent slots going backwards
* because they can actually go backwards if the master crashes when
* it hasn't yet flushed slot state to disk then we copy the older
* slot state after recovery.
*
* There's no checking done for xmin or catalog xmin either, since
* we can't really do anything useful that accounts for xid wrap-around.
*
* Note that this is test harness code. You shouldn't expose slot internals
* to SQL like this for any real world usage. See the README.
*/
Datum
test_slot_timelines_advance_logical_slot(PG_FUNCTION_ARGS)
{
char *slotname = text_to_cstring(PG_GETARG_TEXT_P(0));
TransactionId new_xmin = (TransactionId) PG_GETARG_INT64(1);
TransactionId new_catalog_xmin = (TransactionId) PG_GETARG_INT64(2);
XLogRecPtr restart_lsn = PG_GETARG_LSN(3);
XLogRecPtr confirmed_lsn = PG_GETARG_LSN(4);
CheckSlotRequirements();
ReplicationSlotAcquire(slotname);
if (MyReplicationSlot->data.database != MyDatabaseId)
elog(ERROR, "Trying to update a slot on a different database");
MyReplicationSlot->data.xmin = new_xmin;
MyReplicationSlot->data.catalog_xmin = new_catalog_xmin;
MyReplicationSlot->data.restart_lsn = restart_lsn;
MyReplicationSlot->data.confirmed_flush = confirmed_lsn;
clear_slot_transient_state();
ReplicationSlotMarkDirty();
ReplicationSlotSave();
ReplicationSlotRelease();
ReplicationSlotsComputeRequiredXmin(false);
ReplicationSlotsComputeRequiredLSN();
PG_RETURN_VOID();
}
/*
* SQL function for creating a new physical (streaming replication)
* replication slot.
*/
Datum
pg_create_physical_replication_slot(PG_FUNCTION_ARGS)
{
Name name = PG_GETARG_NAME(0);
Datum values[2];
bool nulls[2];
TupleDesc tupdesc;
HeapTuple tuple;
Datum result;
Assert(!MyReplicationSlot);
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
check_permissions();
CheckSlotRequirements();
/* acquire replication slot, this will check for conflicting names */
ReplicationSlotCreate(NameStr(*name), false, RS_PERSISTENT);
values[0] = NameGetDatum(&MyReplicationSlot->data.name);
nulls[0] = false;
nulls[1] = true;
tuple = heap_form_tuple(tupdesc, values, nulls);
result = HeapTupleGetDatum(tuple);
ReplicationSlotRelease();
PG_RETURN_DATUM(result);
}
/*
* SQL function for creating a new logical replication slot.
*/
Datum
pg_create_logical_replication_slot(PG_FUNCTION_ARGS)
{
Name name = PG_GETARG_NAME(0);
Name plugin = PG_GETARG_NAME(1);
LogicalDecodingContext *ctx = NULL;
TupleDesc tupdesc;
HeapTuple tuple;
Datum result;
Datum values[2];
bool nulls[2];
Assert(!MyReplicationSlot);
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
check_permissions();
CheckLogicalDecodingRequirements();
/*
* Acquire a logical decoding slot, this will check for conflicting names.
* Initially create it as ephemeral - that allows us to nicely handle
* errors during initialization because it'll get dropped if this
* transaction fails. We'll make it persistent at the end.
*/
ReplicationSlotCreate(NameStr(*name), true, RS_EPHEMERAL);
/*
* Create logical decoding context, to build the initial snapshot.
*/
ctx = CreateInitDecodingContext(
NameStr(*plugin), NIL,
logical_read_local_xlog_page, NULL, NULL);
/* build initial snapshot, might take a while */
DecodingContextFindStartpoint(ctx);
values[0] = CStringGetTextDatum(NameStr(MyReplicationSlot->data.name));
values[1] = LSNGetDatum(MyReplicationSlot->data.confirmed_flush);
/* don't need the decoding context anymore */
FreeDecodingContext(ctx);
memset(nulls, 0, sizeof(nulls));
tuple = heap_form_tuple(tupdesc, values, nulls);
result = HeapTupleGetDatum(tuple);
/* ok, slot is now fully created, mark it as persistent */
ReplicationSlotPersist();
ReplicationSlotRelease();
PG_RETURN_DATUM(result);
}
/*
* SQL function for creating a new physical (streaming replication)
* replication slot.
*/
Datum
pg_create_physical_replication_slot(PG_FUNCTION_ARGS)
{
Name name = PG_GETARG_NAME(0);
bool immediately_reserve = PG_GETARG_BOOL(1);
Datum values[2];
bool nulls[2];
TupleDesc tupdesc;
HeapTuple tuple;
Datum result;
Assert(!MyReplicationSlot);
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
check_permissions();
CheckSlotRequirements();
/* acquire replication slot, this will check for conflicting names */
ReplicationSlotCreate(NameStr(*name), false, RS_PERSISTENT);
values[0] = NameGetDatum(&MyReplicationSlot->data.name);
nulls[0] = false;
if (immediately_reserve)
{
/* Reserve WAL as the user asked for it */
ReplicationSlotReserveWal();
/* Write this slot to disk */
ReplicationSlotMarkDirty();
ReplicationSlotSave();
values[1] = LSNGetDatum(MyReplicationSlot->data.restart_lsn);
nulls[1] = false;
}
else
{
values[0] = NameGetDatum(&MyReplicationSlot->data.name);
nulls[1] = true;
}
tuple = heap_form_tuple(tupdesc, values, nulls);
result = HeapTupleGetDatum(tuple);
ReplicationSlotRelease();
PG_RETURN_DATUM(result);
}
/*
* ProcKill() -- Destroy the per-proc data structure for
* this process. Release any of its held LW locks.
*/
static void
ProcKill(int code, Datum arg)
{
/* use volatile pointer to prevent code rearrangement */
volatile PROC_HDR *procglobal = ProcGlobal;
PGPROC *proc;
Assert(MyProc != NULL);
/* Make sure we're out of the sync rep lists */
SyncRepCleanupAtProcExit();
#ifdef USE_ASSERT_CHECKING
{
int i;
/* Last process should have released all locks. */
for (i = 0; i < NUM_LOCK_PARTITIONS; i++)
Assert(SHMQueueEmpty(&(MyProc->myProcLocks[i])));
}
#endif
/*
* Release any LW locks I am holding. There really shouldn't be any, but
* it's cheap to check again before we cut the knees off the LWLock
* facility by releasing our PGPROC ...
*/
LWLockReleaseAll();
/* Make sure active replication slots are released */
if (MyReplicationSlot != NULL)
ReplicationSlotRelease();
/*
* Reset MyLatch to the process local one. This is so that signal
* handlers et al can continue using the latch after the shared latch
* isn't ours anymore. After that clear MyProc and disown the shared
* latch.
*/
SwitchBackToLocalLatch();
proc = MyProc;
MyProc = NULL;
DisownLatch(&proc->procLatch);
SpinLockAcquire(ProcStructLock);
/* Return PGPROC structure (and semaphore) to appropriate freelist */
if (IsAnyAutoVacuumProcess())
{
proc->links.next = (SHM_QUEUE *) procglobal->autovacFreeProcs;
procglobal->autovacFreeProcs = proc;
}
else if (IsBackgroundWorker)
{
proc->links.next = (SHM_QUEUE *) procglobal->bgworkerFreeProcs;
procglobal->bgworkerFreeProcs = proc;
}
else
{
proc->links.next = (SHM_QUEUE *) procglobal->freeProcs;
procglobal->freeProcs = proc;
}
/* Update shared estimate of spins_per_delay */
procglobal->spins_per_delay = update_spins_per_delay(procglobal->spins_per_delay);
SpinLockRelease(ProcStructLock);
/*
* This process is no longer present in shared memory in any meaningful
* way, so tell the postmaster we've cleaned up acceptably well. (XXX
* autovac launcher should be included here someday)
*/
if (IsUnderPostmaster && !IsAutoVacuumLauncherProcess())
MarkPostmasterChildInactive();
/* wake autovac launcher if needed -- see comments in FreeWorkerInfo */
if (AutovacuumLauncherPid != 0)
kill(AutovacuumLauncherPid, SIGUSR2);
}
/*
* Helper function for the various SQL callable logical decoding functions.
*/
static Datum
pg_logical_slot_get_changes_guts(FunctionCallInfo fcinfo, bool confirm, bool binary)
{
Name name;
XLogRecPtr upto_lsn;
int32 upto_nchanges;
ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
MemoryContext per_query_ctx;
MemoryContext oldcontext;
XLogRecPtr end_of_wal;
XLogRecPtr startptr;
LogicalDecodingContext *ctx;
ResourceOwner old_resowner = CurrentResourceOwner;
ArrayType *arr;
Size ndim;
List *options = NIL;
DecodingOutputState *p;
check_permissions();
CheckLogicalDecodingRequirements();
if (PG_ARGISNULL(0))
ereport(ERROR,
(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
errmsg("slot name must not be null")));
name = PG_GETARG_NAME(0);
if (PG_ARGISNULL(1))
upto_lsn = InvalidXLogRecPtr;
else
upto_lsn = PG_GETARG_LSN(1);
if (PG_ARGISNULL(2))
upto_nchanges = InvalidXLogRecPtr;
else
upto_nchanges = PG_GETARG_INT32(2);
if (PG_ARGISNULL(3))
ereport(ERROR,
(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
errmsg("options array must not be null")));
arr = PG_GETARG_ARRAYTYPE_P(3);
/* check to see if caller supports us returning a tuplestore */
if (rsinfo == NULL || !IsA(rsinfo, ReturnSetInfo))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("set-valued function called in context that cannot accept a set")));
if (!(rsinfo->allowedModes & SFRM_Materialize))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("materialize mode required, but it is not allowed in this context")));
/* state to write output to */
p = palloc0(sizeof(DecodingOutputState));
p->binary_output = binary;
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &p->tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
oldcontext = MemoryContextSwitchTo(per_query_ctx);
/* Deconstruct options array */
ndim = ARR_NDIM(arr);
if (ndim > 1)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must be one-dimensional")));
}
else if (array_contains_nulls(arr))
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must not contain nulls")));
}
else if (ndim == 1)
{
int nelems;
Datum *datum_opts;
int i;
Assert(ARR_ELEMTYPE(arr) == TEXTOID);
deconstruct_array(arr, TEXTOID, -1, false, 'i',
&datum_opts, NULL, &nelems);
if (nelems % 2 != 0)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must have even number of elements")));
for (i = 0; i < nelems; i += 2)
{
char *name = TextDatumGetCString(datum_opts[i]);
char *opt = TextDatumGetCString(datum_opts[i + 1]);
options = lappend(options, makeDefElem(name, (Node *) makeString(opt), -1));
}
}
p->tupstore = tuplestore_begin_heap(true, false, work_mem);
rsinfo->returnMode = SFRM_Materialize;
rsinfo->setResult = p->tupstore;
rsinfo->setDesc = p->tupdesc;
/*
* Compute the current end-of-wal and maintain ThisTimeLineID.
* RecoveryInProgress() will update ThisTimeLineID on promotion.
*/
if (!RecoveryInProgress())
end_of_wal = GetFlushRecPtr();
else
end_of_wal = GetXLogReplayRecPtr(&ThisTimeLineID);
ReplicationSlotAcquire(NameStr(*name), true);
PG_TRY();
{
/* restart at slot's confirmed_flush */
ctx = CreateDecodingContext(InvalidXLogRecPtr,
options,
false,
logical_read_local_xlog_page,
LogicalOutputPrepareWrite,
LogicalOutputWrite, NULL);
MemoryContextSwitchTo(oldcontext);
/*
* Check whether the output plugin writes textual output if that's
* what we need.
*/
if (!binary &&
ctx->options.output_type !=OUTPUT_PLUGIN_TEXTUAL_OUTPUT)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("logical decoding output plugin \"%s\" produces binary output, but function \"%s\" expects textual data",
NameStr(MyReplicationSlot->data.plugin),
format_procedure(fcinfo->flinfo->fn_oid))));
ctx->output_writer_private = p;
/*
* Decoding of WAL must start at restart_lsn so that the entirety of
* xacts that committed after the slot's confirmed_flush can be
* accumulated into reorder buffers.
*/
startptr = MyReplicationSlot->data.restart_lsn;
CurrentResourceOwner = ResourceOwnerCreate(CurrentResourceOwner, "logical decoding");
/* invalidate non-timetravel entries */
InvalidateSystemCaches();
/* Decode until we run out of records */
while ((startptr != InvalidXLogRecPtr && startptr < end_of_wal) ||
(ctx->reader->EndRecPtr != InvalidXLogRecPtr && ctx->reader->EndRecPtr < end_of_wal))
{
XLogRecord *record;
char *errm = NULL;
record = XLogReadRecord(ctx->reader, startptr, &errm);
if (errm)
elog(ERROR, "%s", errm);
/*
* Now that we've set up the xlog reader state, subsequent calls
* pass InvalidXLogRecPtr to say "continue from last record"
*/
startptr = InvalidXLogRecPtr;
/*
* The {begin_txn,change,commit_txn}_wrapper callbacks above will
* store the description into our tuplestore.
*/
if (record != NULL)
LogicalDecodingProcessRecord(ctx, ctx->reader);
/* check limits */
if (upto_lsn != InvalidXLogRecPtr &&
upto_lsn <= ctx->reader->EndRecPtr)
break;
if (upto_nchanges != 0 &&
upto_nchanges <= p->returned_rows)
break;
CHECK_FOR_INTERRUPTS();
}
tuplestore_donestoring(tupstore);
CurrentResourceOwner = old_resowner;
/*
* Next time, start where we left off. (Hunting things, the family
* business..)
*/
if (ctx->reader->EndRecPtr != InvalidXLogRecPtr && confirm)
{
LogicalConfirmReceivedLocation(ctx->reader->EndRecPtr);
/*
* If only the confirmed_flush_lsn has changed the slot won't get
* marked as dirty by the above. Callers on the walsender
* interface are expected to keep track of their own progress and
* don't need it written out. But SQL-interface users cannot
* specify their own start positions and it's harder for them to
* keep track of their progress, so we should make more of an
* effort to save it for them.
*
* Dirty the slot so it's written out at the next checkpoint.
* We'll still lose its position on crash, as documented, but it's
* better than always losing the position even on clean restart.
*/
ReplicationSlotMarkDirty();
}
/* free context, call shutdown callback */
FreeDecodingContext(ctx);
ReplicationSlotRelease();
InvalidateSystemCaches();
}
PG_CATCH();
{
/* clear all timetravel entries */
InvalidateSystemCaches();
PG_RE_THROW();
}
PG_END_TRY();
return (Datum) 0;
}
/*
* Helper function for the various SQL callable logical decoding functions.
*/
static Datum
pg_logical_slot_get_changes_guts(FunctionCallInfo fcinfo, bool confirm, bool binary)
{
Name name = PG_GETARG_NAME(0);
XLogRecPtr upto_lsn;
int32 upto_nchanges;
ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
MemoryContext per_query_ctx;
MemoryContext oldcontext;
XLogRecPtr end_of_wal;
XLogRecPtr startptr;
LogicalDecodingContext *ctx;
ResourceOwner old_resowner = CurrentResourceOwner;
ArrayType *arr;
Size ndim;
List *options = NIL;
DecodingOutputState *p;
if (PG_ARGISNULL(1))
upto_lsn = InvalidXLogRecPtr;
else
upto_lsn = PG_GETARG_LSN(1);
if (PG_ARGISNULL(2))
upto_nchanges = InvalidXLogRecPtr;
else
upto_nchanges = PG_GETARG_INT32(2);
/* check to see if caller supports us returning a tuplestore */
if (rsinfo == NULL || !IsA(rsinfo, ReturnSetInfo))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("set-valued function called in context that cannot accept a set")));
if (!(rsinfo->allowedModes & SFRM_Materialize))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("materialize mode required, but it is not allowed in this context")));
/* state to write output to */
p = palloc0(sizeof(DecodingOutputState));
p->binary_output = binary;
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &p->tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
check_permissions();
CheckLogicalDecodingRequirements();
arr = PG_GETARG_ARRAYTYPE_P(3);
ndim = ARR_NDIM(arr);
per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
oldcontext = MemoryContextSwitchTo(per_query_ctx);
if (ndim > 1)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must be one-dimensional")));
}
else if (array_contains_nulls(arr))
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must not contain nulls")));
}
else if (ndim == 1)
{
int nelems;
Datum *datum_opts;
int i;
Assert(ARR_ELEMTYPE(arr) == TEXTOID);
deconstruct_array(arr, TEXTOID, -1, false, 'i',
&datum_opts, NULL, &nelems);
if (nelems % 2 != 0)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("array must have even number of elements")));
for (i = 0; i < nelems; i += 2)
{
char *name = TextDatumGetCString(datum_opts[i]);
char *opt = TextDatumGetCString(datum_opts[i + 1]);
options = lappend(options, makeDefElem(name, (Node *) makeString(opt)));
}
}
p->tupstore = tuplestore_begin_heap(true, false, work_mem);
rsinfo->returnMode = SFRM_Materialize;
rsinfo->setResult = p->tupstore;
rsinfo->setDesc = p->tupdesc;
/* compute the current end-of-wal */
if (!RecoveryInProgress())
end_of_wal = GetFlushRecPtr();
else
end_of_wal = GetXLogReplayRecPtr(NULL);
CheckLogicalDecodingRequirements();
ReplicationSlotAcquire(NameStr(*name));
PG_TRY();
{
ctx = CreateDecodingContext(InvalidXLogRecPtr,
options,
logical_read_local_xlog_page,
LogicalOutputPrepareWrite,
LogicalOutputWrite);
MemoryContextSwitchTo(oldcontext);
/*
* Check whether the output pluggin writes textual output if that's
* what we need.
*/
if (!binary &&
ctx->options.output_type != OUTPUT_PLUGIN_TEXTUAL_OUTPUT)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("output plugin cannot produce binary output")));
ctx->output_writer_private = p;
startptr = MyReplicationSlot->data.restart_lsn;
CurrentResourceOwner = ResourceOwnerCreate(CurrentResourceOwner, "logical decoding");
/* invalidate non-timetravel entries */
InvalidateSystemCaches();
while ((startptr != InvalidXLogRecPtr && startptr < end_of_wal) ||
(ctx->reader->EndRecPtr && ctx->reader->EndRecPtr < end_of_wal))
{
XLogRecord *record;
char *errm = NULL;
record = XLogReadRecord(ctx->reader, startptr, &errm);
if (errm)
elog(ERROR, "%s", errm);
startptr = InvalidXLogRecPtr;
/*
* The {begin_txn,change,commit_txn}_wrapper callbacks above will
* store the description into our tuplestore.
*/
if (record != NULL)
LogicalDecodingProcessRecord(ctx, record);
/* check limits */
if (upto_lsn != InvalidXLogRecPtr &&
upto_lsn <= ctx->reader->EndRecPtr)
break;
if (upto_nchanges != 0 &&
upto_nchanges <= p->returned_rows)
break;
}
}
PG_CATCH();
{
/* clear all timetravel entries */
InvalidateSystemCaches();
PG_RE_THROW();
}
PG_END_TRY();
tuplestore_donestoring(tupstore);
CurrentResourceOwner = old_resowner;
/*
* Next time, start where we left off. (Hunting things, the family
* business..)
*/
if (ctx->reader->EndRecPtr != InvalidXLogRecPtr && confirm)
LogicalConfirmReceivedLocation(ctx->reader->EndRecPtr);
/* free context, call shutdown callback */
FreeDecodingContext(ctx);
ReplicationSlotRelease();
InvalidateSystemCaches();
return (Datum) 0;
}
