/*
* Dynamically launch an SPI worker.
*/
Datum
worker_spi_launch(PG_FUNCTION_ARGS)
{
int32 i = PG_GETARG_INT32(0);
BackgroundWorker worker;
BackgroundWorkerHandle *handle;
BgwHandleStatus status;
pid_t pid;
memset(&worker, 0, sizeof(worker));
worker.bgw_flags = BGWORKER_SHMEM_ACCESS |
BGWORKER_BACKEND_DATABASE_CONNECTION;
worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
worker.bgw_restart_time = BGW_NEVER_RESTART;
sprintf(worker.bgw_library_name, "worker_spi");
sprintf(worker.bgw_function_name, "worker_spi_main");
snprintf(worker.bgw_name, BGW_MAXLEN, "worker_spi worker %d", i);
snprintf(worker.bgw_type, BGW_MAXLEN, "worker_spi");
worker.bgw_main_arg = Int32GetDatum(i);
/* set bgw_notify_pid so that we can use WaitForBackgroundWorkerStartup */
worker.bgw_notify_pid = MyProcPid;
if (!RegisterDynamicBackgroundWorker(&worker, &handle))
PG_RETURN_NULL();
status = WaitForBackgroundWorkerStartup(handle, &pid);
if (status == BGWH_STOPPED)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("could not start background process"),
errhint("More details may be available in the server log.")));
if (status == BGWH_POSTMASTER_DIED)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("cannot start background processes without postmaster"),
errhint("Kill all remaining database processes and restart the database.")));
Assert(status == BGWH_STARTED);
PG_RETURN_INT32(pid);
}
void cfs_start_background_gc()
{
int i;
cfs_state->max_iterations = INT_MAX;
cfs_state->n_workers = cfs_gc_workers;
for (i = 0; i < cfs_gc_workers; i++) {
BackgroundWorker worker;
BackgroundWorkerHandle* handle;
MemSet(&worker, 0, sizeof(worker));
sprintf(worker.bgw_name, "cfs-worker-%d", i);
worker.bgw_flags = BGWORKER_SHMEM_ACCESS;
worker.bgw_start_time = BgWorkerStart_ConsistentState;
worker.bgw_restart_time = 1;
worker.bgw_main = cfs_bgworker_main;
worker.bgw_main_arg = Int32GetDatum(i);
if (!RegisterDynamicBackgroundWorker(&worker, &handle)) {
break;
}
}
elog(LOG, "Start %d background CFS background workers", i);
}
Datum cfs_start_gc(PG_FUNCTION_ARGS)
{
int i = 0;
if (cfs_gc_workers == 0 && pg_atomic_test_set_flag(&cfs_state->gc_started))
{
int j;
BackgroundWorkerHandle** handles;
cfs_stop = true; /* do just one iteration */
cfs_state->max_iterations = 1;
cfs_state->n_workers = PG_GETARG_INT32(0);
handles = (BackgroundWorkerHandle**)palloc(cfs_state->n_workers*sizeof(BackgroundWorkerHandle*));
for (i = 0; i < cfs_state->n_workers; i++) {
BackgroundWorker worker;
MemSet(&worker, 0, sizeof(worker));
sprintf(worker.bgw_name, "cfs-worker-%d", i);
worker.bgw_flags = BGWORKER_SHMEM_ACCESS;
worker.bgw_start_time = BgWorkerStart_ConsistentState;
worker.bgw_restart_time = 1;
worker.bgw_main = cfs_bgworker_main;
worker.bgw_main_arg = Int32GetDatum(i);
if (!RegisterDynamicBackgroundWorker(&worker, &handles[i])) {
break;
}
}
for (j = 0; j < i; j++) {
WaitForBackgroundWorkerShutdown(handles[j]);
}
pfree(handles);
pg_atomic_clear_flag(&cfs_state->gc_started);
}
PG_RETURN_INT32(i);
}
static bool
run_background_proc(ContQueryProc *proc)
{
BackgroundWorker worker;
BackgroundWorkerHandle *handle;
bool success;
strcpy(worker.bgw_name, GetContQueryProcName(proc));
worker.bgw_flags = BGWORKER_SHMEM_ACCESS | BGWORKER_BACKEND_DATABASE_CONNECTION;
worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
worker.bgw_main = cq_bgproc_main;
worker.bgw_notify_pid = MyProcPid;
worker.bgw_restart_time = 1; /* recover in 1s */
worker.bgw_let_crash = false;
worker.bgw_main_arg = PointerGetDatum(proc);
success = RegisterDynamicBackgroundWorker(&worker, &handle);
if (success)
proc->handle = *handle;
return success;
}
/*
* launch_consumer_group
*
* Launch a group of background worker process that will consume from the given topic
* into the given relation
*/
static bool
launch_consumer_group(Relation consumers, KafkaConsumer *consumer, int64 offset)
{
BackgroundWorker worker;
BackgroundWorkerHandle *handle;
KafkaConsumerGroup *group;
bool found;
int i;
group = (KafkaConsumerGroup *) hash_search(consumer_groups, &consumer->id, HASH_ENTER, &found);
if (found)
{
KafkaConsumerProc *proc;
HASH_SEQ_STATUS iter;
bool running = false;
hash_seq_init(&iter, consumer_procs);
while ((proc = (KafkaConsumerProc *) hash_seq_search(&iter)) != NULL)
{
if (proc->consumer_id == consumer->id)
{
running = true;
break;
}
}
hash_seq_term(&iter);
/* if there are already procs running, it's a noop */
if (running)
return true;
/* no procs actually running, so it's ok to launch new ones */
}
group->parallelism = consumer->parallelism;
for (i = 0; i < group->parallelism; i++)
{
/* we just need any unique OID here */
Oid id = GetNewOid(consumers);
KafkaConsumerProc *proc;
proc = (KafkaConsumerProc *) hash_search(consumer_procs, &id, HASH_ENTER, &found);
if (found)
continue;
worker.bgw_main_arg = DatumGetObjectId(id);
worker.bgw_flags = BGWORKER_BACKEND_DATABASE_CONNECTION | BGWORKER_SHMEM_ACCESS;
worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
worker.bgw_restart_time = BGW_NEVER_RESTART;
worker.bgw_main = NULL;
worker.bgw_notify_pid = 0;
/* this module is loaded dynamically, so we can't use bgw_main */
sprintf(worker.bgw_library_name, PIPELINE_KAFKA_LIB);
sprintf(worker.bgw_function_name, KAFKA_CONSUME_MAIN);
snprintf(worker.bgw_name, BGW_MAXLEN, "[kafka consumer] %s <- %s", consumer->rel->relname, consumer->topic);
proc->consumer_id = consumer->id;
proc->partition_group = i;
proc->offset = offset;
namestrcpy(&proc->dbname, get_database_name(MyDatabaseId));
if (!RegisterDynamicBackgroundWorker(&worker, &handle))
return false;
proc->worker = *handle;
}
return true;
}
/*
* Start new apply background worker, if possible.
*/
void
logicalrep_worker_launch(Oid dbid, Oid subid, const char *subname, Oid userid,
Oid relid)
{
BackgroundWorker bgw;
BackgroundWorkerHandle *bgw_handle;
uint16 generation;
int i;
int slot = 0;
LogicalRepWorker *worker = NULL;
int nsyncworkers;
TimestampTz now;
ereport(DEBUG1,
(errmsg("starting logical replication worker for subscription \"%s\"",
subname)));
/* Report this after the initial starting message for consistency. */
if (max_replication_slots == 0)
ereport(ERROR,
(errcode(ERRCODE_CONFIGURATION_LIMIT_EXCEEDED),
errmsg("cannot start logical replication workers when max_replication_slots = 0")));
/*
* We need to do the modification of the shared memory under lock so that
* we have consistent view.
*/
LWLockAcquire(LogicalRepWorkerLock, LW_EXCLUSIVE);
retry:
/* Find unused worker slot. */
for (i = 0; i < max_logical_replication_workers; i++)
{
LogicalRepWorker *w = &LogicalRepCtx->workers[i];
if (!w->in_use)
{
worker = w;
slot = i;
break;
}
}
nsyncworkers = logicalrep_sync_worker_count(subid);
now = GetCurrentTimestamp();
/*
* If we didn't find a free slot, try to do garbage collection. The
* reason we do this is because if some worker failed to start up and its
* parent has crashed while waiting, the in_use state was never cleared.
*/
if (worker == NULL || nsyncworkers >= max_sync_workers_per_subscription)
{
bool did_cleanup = false;
for (i = 0; i < max_logical_replication_workers; i++)
{
LogicalRepWorker *w = &LogicalRepCtx->workers[i];
/*
* If the worker was marked in use but didn't manage to attach in
* time, clean it up.
*/
if (w->in_use && !w->proc &&
TimestampDifferenceExceeds(w->launch_time, now,
wal_receiver_timeout))
{
elog(WARNING,
"logical replication worker for subscription %u took too long to start; canceled",
w->subid);
logicalrep_worker_cleanup(w);
did_cleanup = true;
}
}
if (did_cleanup)
goto retry;
}
/*
* If we reached the sync worker limit per subscription, just exit
* silently as we might get here because of an otherwise harmless race
* condition.
*/
if (nsyncworkers >= max_sync_workers_per_subscription)
{
LWLockRelease(LogicalRepWorkerLock);
return;
}
/*
* However if there are no more free worker slots, inform user about it
* before exiting.
*/
if (worker == NULL)
{
LWLockRelease(LogicalRepWorkerLock);
ereport(WARNING,
(errcode(ERRCODE_CONFIGURATION_LIMIT_EXCEEDED),
errmsg("out of logical replication worker slots"),
errhint("You might need to increase max_logical_replication_workers.")));
return;
}
/* Prepare the worker slot. */
worker->launch_time = now;
worker->in_use = true;
worker->generation++;
worker->proc = NULL;
worker->dbid = dbid;
worker->userid = userid;
worker->subid = subid;
worker->relid = relid;
worker->relstate = SUBREL_STATE_UNKNOWN;
worker->relstate_lsn = InvalidXLogRecPtr;
worker->last_lsn = InvalidXLogRecPtr;
TIMESTAMP_NOBEGIN(worker->last_send_time);
TIMESTAMP_NOBEGIN(worker->last_recv_time);
worker->reply_lsn = InvalidXLogRecPtr;
TIMESTAMP_NOBEGIN(worker->reply_time);
/* Before releasing lock, remember generation for future identification. */
generation = worker->generation;
LWLockRelease(LogicalRepWorkerLock);
/* Register the new dynamic worker. */
memset(&bgw, 0, sizeof(bgw));
bgw.bgw_flags = BGWORKER_SHMEM_ACCESS |
BGWORKER_BACKEND_DATABASE_CONNECTION;
bgw.bgw_start_time = BgWorkerStart_RecoveryFinished;
snprintf(bgw.bgw_library_name, BGW_MAXLEN, "postgres");
snprintf(bgw.bgw_function_name, BGW_MAXLEN, "ApplyWorkerMain");
if (OidIsValid(relid))
snprintf(bgw.bgw_name, BGW_MAXLEN,
"logical replication worker for subscription %u sync %u", subid, relid);
else
snprintf(bgw.bgw_name, BGW_MAXLEN,
"logical replication worker for subscription %u", subid);
snprintf(bgw.bgw_type, BGW_MAXLEN, "logical replication worker");
bgw.bgw_restart_time = BGW_NEVER_RESTART;
bgw.bgw_notify_pid = MyProcPid;
bgw.bgw_main_arg = Int32GetDatum(slot);
if (!RegisterDynamicBackgroundWorker(&bgw, &bgw_handle))
{
/* Failed to start worker, so clean up the worker slot. */
LWLockAcquire(LogicalRepWorkerLock, LW_EXCLUSIVE);
Assert(generation == worker->generation);
logicalrep_worker_cleanup(worker);
LWLockRelease(LogicalRepWorkerLock);
ereport(WARNING,
(errcode(ERRCODE_CONFIGURATION_LIMIT_EXCEEDED),
errmsg("out of background worker slots"),
errhint("You might need to increase max_worker_processes.")));
return;
}
/* Now wait until it attaches. */
WaitForReplicationWorkerAttach(worker, generation, bgw_handle);
}
/*
* Register background workers.
*/
static worker_state *
setup_background_workers(int nworkers, dsm_segment *seg)
{
MemoryContext oldcontext;
BackgroundWorker worker;
worker_state *wstate;
int i;
/*
* We need the worker_state object and the background worker handles to
* which it points to be allocated in CurTransactionContext rather than
* ExprContext; otherwise, they'll be destroyed before the on_dsm_detach
* hooks run.
*/
oldcontext = MemoryContextSwitchTo(CurTransactionContext);
/* Create worker state object. */
wstate = MemoryContextAlloc(TopTransactionContext,
offsetof(worker_state, handle) +
sizeof(BackgroundWorkerHandle *) * nworkers);
wstate->nworkers = 0;
/*
* Arrange to kill all the workers if we abort before all workers are
* finished hooking themselves up to the dynamic shared memory segment.
*
* If we die after all the workers have finished hooking themselves up to
* the dynamic shared memory segment, we'll mark the two queues to which
* we're directly connected as detached, and the worker(s) connected to
* those queues will exit, marking any other queues to which they are
* connected as detached. This will cause any as-yet-unaware workers
* connected to those queues to exit in their turn, and so on, until
* everybody exits.
*
* But suppose the workers which are supposed to connect to the queues to
* which we're directly attached exit due to some error before they
* actually attach the queues. The remaining workers will have no way of
* knowing this. From their perspective, they're still waiting for those
* workers to start, when in fact they've already died.
*/
on_dsm_detach(seg, cleanup_background_workers,
PointerGetDatum(wstate));
/* Configure a worker. */
worker.bgw_flags = BGWORKER_SHMEM_ACCESS;
worker.bgw_start_time = BgWorkerStart_ConsistentState;
worker.bgw_restart_time = BGW_NEVER_RESTART;
worker.bgw_main = NULL; /* new worker might not have library loaded */
sprintf(worker.bgw_library_name, "test_shm_mq");
sprintf(worker.bgw_function_name, "test_shm_mq_main");
snprintf(worker.bgw_name, BGW_MAXLEN, "test_shm_mq");
worker.bgw_main_arg = UInt32GetDatum(dsm_segment_handle(seg));
/* set bgw_notify_pid, so we can detect if the worker stops */
worker.bgw_notify_pid = MyProcPid;
/* Register the workers. */
for (i = 0; i < nworkers; ++i)
{
if (!RegisterDynamicBackgroundWorker(&worker, &wstate->handle[i]))
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("could not register background process"),
errhint("You may need to increase max_worker_processes.")));
++wstate->nworkers;
}
/* All done. */
MemoryContextSwitchTo(oldcontext);
return wstate;
}
Datum
worker_test(PG_FUNCTION_ARGS)
{
int i, nworkers;
dsm_segment *seg;
test_shm_mq_header *hdr;
MemoryContext oldcontext;
worker_state *wstate;
nworkers = PG_GETARG_INT32(0);
#if PG_VERSION_NUM >= 90500
seg = dsm_create(sizeof(test_shm_mq_header), 0);
#else
seg = dsm_create(sizeof(test_shm_mq_header)
#endif
hdr = dsm_segment_address(seg);
printf("begin worker_test: %d, %p\n", dsm_segment_handle(seg), hdr);
MemSet(hdr, 0, sizeof(test_shm_mq_header));
SpinLockInit(&hdr->mutex);
strncpy(hdr->dbname, get_database_name(MyDatabaseId), sizeof(hdr->dbname));
oldcontext = MemoryContextSwitchTo(CurTransactionContext);
wstate = MemoryContextAlloc(TopTransactionContext,
offsetof(worker_state, handle) +
sizeof(BackgroundWorkerHandle *) * nworkers);
MemSet(wstate, 0, offsetof(worker_state, handle) + sizeof(BackgroundWorkerHandle *) * nworkers);
on_dsm_detach(seg, cleanup_background_workers,
PointerGetDatum(wstate));
for (i = 0; i < nworkers; i++)
{
BackgroundWorker worker;
MemSet(&worker, 0, sizeof(worker));
worker.bgw_flags = BGWORKER_SHMEM_ACCESS | BGWORKER_BACKEND_DATABASE_CONNECTION;
worker.bgw_start_time = BgWorkerStart_ConsistentState;
worker.bgw_restart_time = BGW_NEVER_RESTART;
worker.bgw_main = NULL; /* new worker might not have library loaded */
sprintf(worker.bgw_library_name, "worker_test");
sprintf(worker.bgw_function_name, "worker_test_main");
snprintf(worker.bgw_name, BGW_MAXLEN, "worker_test %d", i);
worker.bgw_main_arg = UInt32GetDatum(dsm_segment_handle(seg));
worker.bgw_notify_pid = MyProcPid;
if (!RegisterDynamicBackgroundWorker(&worker, &wstate->handle[i]))
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("could not register background process"),
errhint("You may need to increase max_worker_processes.")));
++wstate->nworkers;
}
for (i = 0; i < nworkers; i++)
{
BgwHandleStatus status;
pid_t pid;
status = WaitForBackgroundWorkerStartup(wstate->handle[i], &pid);
if (status == BGWH_STOPPED)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("could not start background process"),
errhint("More details may be available in the server log.")));
if (status == BGWH_POSTMASTER_DIED)
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("cannot start background processes without postmaster"),
errhint("Kill all remaining database processes and restart the database.")));
Assert(status == BGWH_STARTED);
}
wait_for_workers_to_become_ready(wstate, hdr);
cancel_on_dsm_detach(seg, cleanup_background_workers,
PointerGetDatum(wstate));
dsm_detach(seg);
pfree(wstate);
MemoryContextSwitchTo(oldcontext);
PG_RETURN_VOID();
}
/*
* Starts background worker that will create new partitions,
* waits till it finishes the job and returns the result (new partition oid)
*/
Oid
create_partitions_bg_worker(Oid relid, Datum value, Oid value_type, bool *crashed)
{
BackgroundWorker worker;
BackgroundWorkerHandle *worker_handle;
BgwHandleStatus status;
dsm_segment *segment;
dsm_handle segment_handle;
pid_t pid;
PartitionArgs *args;
Oid child_oid;
TypeCacheEntry *tce;
/* Create a dsm segment for the worker to pass arguments */
segment = dsm_create(sizeof(PartitionArgs), 0);
segment_handle = dsm_segment_handle(segment);
tce = lookup_type_cache(value_type, 0);
/* Fill arguments structure */
args = (PartitionArgs *) dsm_segment_address(segment);
args->dbid = MyDatabaseId;
args->relid = relid;
if (tce->typbyval)
args->value = value;
else
memcpy(&args->value, DatumGetPointer(value), sizeof(args->value));
args->by_val = tce->typbyval;
args->value_type = value_type;
args->result = 0;
/* Initialize worker struct */
worker.bgw_flags = BGWORKER_SHMEM_ACCESS |
BGWORKER_BACKEND_DATABASE_CONNECTION;
worker.bgw_start_time = BgWorkerStart_RecoveryFinished;
worker.bgw_restart_time = BGW_NEVER_RESTART;
worker.bgw_main = bg_worker_main;
worker.bgw_main_arg = Int32GetDatum(segment_handle);
worker.bgw_notify_pid = MyProcPid;
/* Start dynamic worker */
if (!RegisterDynamicBackgroundWorker(&worker, &worker_handle))
{
elog(WARNING, "Unable to create background worker for pg_pathman");
}
status = WaitForBackgroundWorkerStartup(worker_handle, &pid);
if (status == BGWH_POSTMASTER_DIED)
{
ereport(WARNING,
(errmsg("Postmaster died during the pg_pathman background worker process"),
errhint("More details may be available in the server log.")));
}
/* Wait till the worker finishes its job */
status = WaitForBackgroundWorkerShutdown(worker_handle);
*crashed = args->crashed;
child_oid = args->result;
/* Free dsm segment */
dsm_detach(segment);
return child_oid;
}
