int
CommitTranGTM(GlobalTransactionId gxid)
{
int ret;
if (!GlobalTransactionIdIsValid(gxid))
return 0;
CheckConnection();
ret = commit_transaction(conn, gxid);
/*
* If something went wrong (timeout), try and reset GTM connection.
* We will close the transaction locally anyway, and closing GTM will force
* it to be closed on GTM.
*/
if (ret < 0)
{
CloseGTM();
InitGTM();
}
/* Close connection in case commit is done by autovacuum worker or launcher */
if (IsAutoVacuumWorkerProcess() || IsAutoVacuumLauncherProcess())
CloseGTM();
return ret;
}
void
InitGTM(void)
{
/* 256 bytes should be enough */
char conn_str[256];
/* If this thread is postmaster itself, it contacts gtm identifying itself */
if (!IsUnderPostmaster)
{
GTM_PGXCNodeType remote_type = GTM_NODE_DEFAULT;
if (IS_PGXC_COORDINATOR)
remote_type = GTM_NODE_COORDINATOR;
else if (IS_PGXC_DATANODE)
remote_type = GTM_NODE_DATANODE;
sprintf(conn_str, "host=%s port=%d node_name=%s remote_type=%d postmaster=1",
GtmHost, GtmPort, PGXCNodeName, remote_type);
/* Log activity of GTM connections */
elog(DEBUG1, "Postmaster: connection established to GTM with string %s", conn_str);
}
else
{
sprintf(conn_str, "host=%s port=%d node_name=%s", GtmHost, GtmPort, PGXCNodeName);
/* Log activity of GTM connections */
if (IsAutoVacuumWorkerProcess())
elog(DEBUG1, "Autovacuum worker: connection established to GTM with string %s", conn_str);
else if (IsAutoVacuumLauncherProcess())
elog(DEBUG1, "Autovacuum launcher: connection established to GTM with string %s", conn_str);
else
elog(DEBUG1, "Postmaster child: connection established to GTM with string %s", conn_str);
}
conn = PQconnectGTM(conn_str);
if (GTMPQstatus(conn) != CONNECTION_OK)
{
int save_errno = errno;
ereport(WARNING,
(errcode(ERRCODE_CONNECTION_EXCEPTION),
errmsg("can not connect to GTM: %m")));
errno = save_errno;
CloseGTM();
}
#ifdef XCP
else if (IS_PGXC_COORDINATOR)
register_session(conn, PGXCNodeName, MyProcPid, MyBackendId);
#endif
}
/* ----------------------------------------------------------------
* proc_exit
*
* this function calls all the callbacks registered
* for it (to free resources) and then calls exit.
*
* This should be the only function to call exit().
* -cim 2/6/90
*
* Unfortunately, we can't really guarantee that add-on code
* obeys the rule of not calling exit() directly. So, while
* this is the preferred way out of the system, we also register
* an atexit callback that will make sure cleanup happens.
* ----------------------------------------------------------------
*/
void
proc_exit(int code)
{
pqsignal(SIGALRM, SIG_IGN);
/* Clean up everything that must be cleaned up */
proc_exit_prepare(code);
#ifdef PROFILE_PID_DIR
{
/*
* If we are profiling ourself then gprof's mcleanup() is about to
* write out a profile to ./gmon.out. Since mcleanup() always uses a
* fixed file name, each backend will overwrite earlier profiles. To
* fix that, we create a separate subdirectory for each backend
* (./gprof/pid) and 'cd' to that subdirectory before we exit() - that
* forces mcleanup() to write each profile into its own directory. We
* end up with something like: $PGDATA/gprof/8829/gmon.out
* $PGDATA/gprof/8845/gmon.out ...
*
* To avoid undesirable disk space bloat, autovacuum workers are
* discriminated against: all their gmon.out files go into the same
* subdirectory. Without this, an installation that is "just sitting
* there" nonetheless eats megabytes of disk space every few seconds.
*
* Note that we do this here instead of in an on_proc_exit() callback
* because we want to ensure that this code executes last - we don't
* want to interfere with any other on_proc_exit() callback. For the
* same reason, we do not include it in proc_exit_prepare ... so if
* you are exiting in the "wrong way" you won't drop your profile in a
* nice place.
*/
char gprofDirName[32];
if (IsAutoVacuumWorkerProcess())
snprintf(gprofDirName, 32, "gprof/avworker");
else
snprintf(gprofDirName, 32, "gprof/%d", (int) getpid());
mkdir("gprof", S_IRWXU | S_IRWXG | S_IRWXO);
mkdir(gprofDirName, S_IRWXU | S_IRWXG | S_IRWXO);
chdir(gprofDirName);
}
#endif
elog(DEBUG3, "exit(%d)", code);
exit(code);
}
void
CloseGTM(void)
{
GTMPQfinish(conn);
conn = NULL;
/* Log activity of GTM connections */
if (!IsUnderPostmaster)
elog(DEBUG1, "Postmaster: connection to GTM closed");
else if (IsAutoVacuumWorkerProcess())
elog(DEBUG1, "Autovacuum worker: connection to GTM closed");
else if (IsAutoVacuumLauncherProcess())
elog(DEBUG1, "Autovacuum launcher: connection to GTM closed");
else
elog(DEBUG1, "Postmaster child: connection to GTM closed");
}
int
CommitTranGTM(GlobalTransactionId gxid, int waited_xid_count,
GlobalTransactionId *waited_xids)
{
int ret;
if (!GlobalTransactionIdIsValid(gxid))
return 0;
CheckConnection();
#ifdef XCP
ret = -1;
if (conn)
#endif
ret = commit_transaction(conn, gxid, waited_xid_count, waited_xids);
/*
* If something went wrong (timeout), try and reset GTM connection.
* We will close the transaction locally anyway, and closing GTM will force
* it to be closed on GTM.
*/
if (ret < 0)
{
CloseGTM();
InitGTM();
#ifdef XCP
if (conn)
ret = commit_transaction(conn, gxid, waited_xid_count, waited_xids);
#endif
}
/* Close connection in case commit is done by autovacuum worker or launcher */
if (IsAutoVacuumWorkerProcess() || IsAutoVacuumLauncherProcess())
CloseGTM();
currentGxid = InvalidGlobalTransactionId;
return ret;
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* The database can be specified by name, using the in_dbname parameter, or by
* OID, using the dboid parameter. In the latter case, the actual database
* name can be returned to the caller in out_dbname. If out_dbname isn't
* NULL, it must point to a buffer of size NAMEDATALEN.
*
* In bootstrap mode no parameters are used. The autovacuum launcher process
* doesn't use any parameters either, because it only goes far enough to be
* able to read pg_database; it doesn't connect to any particular database.
* In walsender mode only username is used.
*
* As of PostgreSQL 8.2, we expect InitProcess() was already called, so we
* already have a PGPROC struct ... but it's not completely filled in yet.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
void
InitPostgres(const char *in_dbname, Oid dboid, const char *username,
char *out_dbname)
{
bool bootstrap = IsBootstrapProcessingMode();
bool am_superuser;
GucContext gucctx;
char *fullpath;
char dbname[NAMEDATALEN];
elog(DEBUG3, "InitPostgres");
/*
* Add my PGPROC struct to the ProcArray.
*
* Once I have done this, I am visible to other backends!
*/
InitProcessPhase2();
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data.
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
SharedInvalBackendInit(false);
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend id: %d", MyBackendId);
/* Now that we have a BackendId, we can participate in ProcSignal */
ProcSignalInit(MyBackendId);
/*
* bufmgr needs another initialization call too
*/
InitBufferPoolBackend();
/*
* Initialize local process's access to XLOG, if appropriate. In
* bootstrap case we skip this since StartupXLOG() was run instead.
*/
if (!bootstrap)
(void) RecoveryInProgress();
/*
* Initialize the relation cache and the system catalog caches. Note that
* no catalog access happens here; we only set up the hashtable structure.
* We must do this before starting a transaction because transaction abort
* would try to touch these hashtables.
*/
RelationCacheInitialize();
InitCatalogCache();
InitPlanCache();
/* Initialize portal manager */
EnablePortalManager();
/* Initialize stats collection --- must happen before first xact */
if (!bootstrap)
pgstat_initialize();
/*
* Load relcache entries for the shared system catalogs. This must create
* at least an entry for pg_database.
*/
RelationCacheInitializePhase2();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This has to
* be after we've initialized all the low-level modules like the buffer
* manager, because during shutdown this has to run before the low-level
* modules start to close down. On the other hand, we want it in place
* before we begin our first transaction --- if we fail during the
* initialization transaction, as is entirely possible, we need the
* AbortTransaction call to clean up.
*/
on_shmem_exit(ShutdownPostgres, 0);
/* The autovacuum launcher is done here */
if (IsAutoVacuumLauncherProcess())
return;
/*
* Start a new transaction here before first access to db, and get a
* snapshot. We don't have a use for the snapshot itself, but we're
* interested in the secondary effect that it sets RecentGlobalXmin. (This
* is critical for anything that reads heap pages, because HOT may decide
* to prune them even if the process doesn't attempt to modify any
* tuples.)
*/
if (!bootstrap)
{
StartTransactionCommand();
(void) GetTransactionSnapshot();
}
/*
* Set up the global variables holding database id and default tablespace.
* But note we won't actually try to touch the database just yet.
*
* We take a shortcut in the bootstrap and walsender case, otherwise we
* have to look up the db's entry in pg_database.
*/
if (bootstrap || am_walsender)
{
MyDatabaseId = TemplateDbOid;
MyDatabaseTableSpace = DEFAULTTABLESPACE_OID;
}
else if (in_dbname != NULL)
{
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTuple(in_dbname);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", in_dbname)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
/* take database name from the caller, just for paranoia */
strlcpy(dbname, in_dbname, sizeof(dbname));
}
else
{
/* caller specified database by OID */
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTupleByOid(dboid);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database %u does not exist", dboid)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
Assert(MyDatabaseId == dboid);
strlcpy(dbname, NameStr(dbform->datname), sizeof(dbname));
/* pass the database name back to the caller */
if (out_dbname)
strcpy(out_dbname, dbname);
}
/* Now we can mark our PGPROC entry with the database ID */
/* (We assume this is an atomic store so no lock is needed) */
MyProc->databaseId = MyDatabaseId;
/*
* Now, take a writer's lock on the database we are trying to connect to.
* If there is a concurrently running DROP DATABASE on that database, this
* will block us until it finishes (and has committed its update of
* pg_database).
*
* Note that the lock is not held long, only until the end of this startup
* transaction. This is OK since we are already advertising our use of
* the database in the PGPROC array; anyone trying a DROP DATABASE after
* this point will see us there.
*
* Note: use of RowExclusiveLock here is reasonable because we envision
* our session as being a concurrent writer of the database. If we had a
* way of declaring a session as being guaranteed-read-only, we could use
* AccessShareLock for such sessions and thereby not conflict against
* CREATE DATABASE.
*/
if (!bootstrap && !am_walsender)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
RowExclusiveLock);
/*
* Recheck pg_database to make sure the target database hasn't gone away.
* If there was a concurrent DROP DATABASE, this ensures we will die
* cleanly without creating a mess.
*/
if (!bootstrap && !am_walsender)
{
HeapTuple tuple;
tuple = GetDatabaseTuple(dbname);
if (!HeapTupleIsValid(tuple) ||
MyDatabaseId != HeapTupleGetOid(tuple) ||
MyDatabaseTableSpace != ((Form_pg_database) GETSTRUCT(tuple))->dattablespace)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", dbname),
errdetail("It seems to have just been dropped or renamed.")));
}
/*
* Now we should be able to access the database directory safely. Verify
* it's there and looks reasonable.
*/
fullpath = GetDatabasePath(MyDatabaseId, MyDatabaseTableSpace);
if (!bootstrap && !am_walsender)
{
if (access(fullpath, F_OK) == -1)
{
if (errno == ENOENT)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("The database subdirectory \"%s\" is missing.",
fullpath)));
else
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not access directory \"%s\": %m",
fullpath)));
}
ValidatePgVersion(fullpath);
}
SetDatabasePath(fullpath);
/*
* It's now possible to do real access to the system catalogs.
*
* Load relcache entries for the system catalogs. This must create at
* least the minimum set of "nailed-in" cache entries.
*/
RelationCacheInitializePhase3();
/*
* Perform client authentication if necessary, then figure out our
* postgres user ID, and see if we are a superuser.
*
* In standalone mode and in autovacuum worker processes, we use a fixed
* ID, otherwise we figure it out from the authenticated user name.
*/
if (bootstrap || IsAutoVacuumWorkerProcess())
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
if (!ThereIsAtLeastOneRole())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no roles are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" SUPERUSER;.",
username)));
}
else
{
/* normal multiuser case */
Assert(MyProcPort != NULL);
PerformAuthentication(MyProcPort);
InitializeSessionUserId(username);
am_superuser = superuser();
}
/* set up ACL framework (so CheckMyDatabase can check permissions) */
initialize_acl();
/* Process pg_db_role_setting options */
process_settings(MyDatabaseId, GetSessionUserId());
/*
* Re-read the pg_database row for our database, check permissions and set
* up database-specific GUC settings. We can't do this until all the
* database-access infrastructure is up. (Also, it wants to know if the
* user is a superuser, so the above stuff has to happen first.)
*/
if (!bootstrap && !am_walsender)
CheckMyDatabase(dbname, am_superuser);
/*
* If we're trying to shut down, only superusers can connect.
*/
if (!am_superuser &&
MyProcPort != NULL &&
MyProcPort->canAcceptConnections == CAC_WAITBACKUP)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to connect during database shutdown")));
/*
* Check a normal user hasn't connected to a superuser reserved slot.
*/
if (!am_superuser &&
ReservedBackends > 0 &&
!HaveNFreeProcs(ReservedBackends))
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("connection limit exceeded for non-superusers")));
/*
* Now process any command-line switches that were included in the startup
* packet, if we are in a regular backend. We couldn't do this before
* because we didn't know if client is a superuser.
*/
gucctx = am_superuser ? PGC_SUSET : PGC_BACKEND;
if (MyProcPort != NULL &&
MyProcPort->cmdline_options != NULL)
{
/*
* The maximum possible number of commandline arguments that could
* come from MyProcPort->cmdline_options is (strlen + 1) / 2; see
* pg_split_opts().
*/
char **av;
int maxac;
int ac;
maxac = 2 + (strlen(MyProcPort->cmdline_options) + 1) / 2;
av = (char **) palloc(maxac * sizeof(char *));
ac = 0;
av[ac++] = "postgres";
/* Note this mangles MyProcPort->cmdline_options */
pg_split_opts(av, &ac, MyProcPort->cmdline_options);
av[ac] = NULL;
Assert(ac < maxac);
(void) process_postgres_switches(ac, av, gucctx);
}
/*
* Process any additional GUC variable settings passed in startup packet.
* These are handled exactly like command-line variables.
*/
if (MyProcPort != NULL)
{
ListCell *gucopts = list_head(MyProcPort->guc_options);
while (gucopts)
{
char *name;
char *value;
name = lfirst(gucopts);
gucopts = lnext(gucopts);
value = lfirst(gucopts);
gucopts = lnext(gucopts);
SetConfigOption(name, value, gucctx, PGC_S_CLIENT);
}
}
/* Apply PostAuthDelay as soon as we've read all options */
if (PostAuthDelay > 0)
pg_usleep(PostAuthDelay * 1000000L);
/*
* Initialize various default states that can't be set up until we've
* selected the active user and gotten the right GUC settings.
*/
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/* reset the database for walsender */
if (am_walsender)
MyProc->databaseId = MyDatabaseId = InvalidOid;
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* The database can be specified by name, using the in_dbname parameter, or by
* OID, using the dboid parameter. In the latter case, the computed database
* name is passed out to the caller as a palloc'ed string in out_dbname.
*
* In bootstrap mode no parameters are used.
*
* The return value indicates whether the userID is a superuser. (That
* can only be tested inside a transaction, so we want to do it during
* the startup transaction rather than doing a separate one in postgres.c.)
*
* As of PostgreSQL 8.2, we expect InitProcess() was already called, so we
* already have a PGPROC struct ... but it's not filled in yet.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
bool
InitPostgres(const char *in_dbname, Oid dboid, const char *username,
char **out_dbname)
{
bool bootstrap = IsBootstrapProcessingMode();
bool autovacuum = IsAutoVacuumWorkerProcess();
bool am_superuser;
char *fullpath;
char dbname[NAMEDATALEN];
/*
* Set up the global variables holding database id and path. But note we
* won't actually try to touch the database just yet.
*
* We take a shortcut in the bootstrap case, otherwise we have to look up
* the db name in pg_database.
*/
if (bootstrap)
{
MyDatabaseId = TemplateDbOid;
MyDatabaseTableSpace = DEFAULTTABLESPACE_OID;
}
else
{
/*
* Find tablespace of the database we're about to open. Since we're
* not yet up and running we have to use one of the hackish
* FindMyDatabase variants, which look in the flat-file copy of
* pg_database.
*
* If the in_dbname param is NULL, lookup database by OID.
*/
if (in_dbname == NULL)
{
if (!FindMyDatabaseByOid(dboid, dbname, &MyDatabaseTableSpace))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database %u does not exist", dboid)));
MyDatabaseId = dboid;
/* pass the database name to the caller */
*out_dbname = pstrdup(dbname);
}
else
{
if (!FindMyDatabase(in_dbname, &MyDatabaseId, &MyDatabaseTableSpace))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
in_dbname)));
/* our database name is gotten from the caller */
strlcpy(dbname, in_dbname, NAMEDATALEN);
}
}
fullpath = GetDatabasePath(MyDatabaseId, MyDatabaseTableSpace);
SetDatabasePath(fullpath);
/*
* Finish filling in the PGPROC struct, and add it to the ProcArray. (We
* need to know MyDatabaseId before we can do this, since it's entered
* into the PGPROC struct.)
*
* Once I have done this, I am visible to other backends!
*/
InitProcessPhase2();
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data.
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
SharedInvalBackendInit();
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend id: %d", MyBackendId);
/*
* bufmgr needs another initialization call too
*/
InitBufferPoolBackend();
/*
* Initialize local process's access to XLOG. In bootstrap case we may
* skip this since StartupXLOG() was run instead.
*/
if (!bootstrap)
InitXLOGAccess();
/*
* Initialize the relation cache and the system catalog caches. Note that
* no catalog access happens here; we only set up the hashtable structure.
* We must do this before starting a transaction because transaction abort
* would try to touch these hashtables.
*/
RelationCacheInitialize();
InitCatalogCache();
InitPlanCache();
/* Initialize portal manager */
EnablePortalManager();
/* Initialize stats collection --- must happen before first xact */
if (!bootstrap)
pgstat_initialize();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This has to
* be after we've initialized all the low-level modules like the buffer
* manager, because during shutdown this has to run before the low-level
* modules start to close down. On the other hand, we want it in place
* before we begin our first transaction --- if we fail during the
* initialization transaction, as is entirely possible, we need the
* AbortTransaction call to clean up.
*/
on_shmem_exit(ShutdownPostgres, 0);
/*
* Start a new transaction here before first access to db, and get a
* snapshot. We don't have a use for the snapshot itself, but we're
* interested in the secondary effect that it sets RecentGlobalXmin.
*/
if (!bootstrap)
{
StartTransactionCommand();
(void) GetTransactionSnapshot();
}
/*
* Now that we have a transaction, we can take locks. Take a writer's
* lock on the database we are trying to connect to. If there is a
* concurrently running DROP DATABASE on that database, this will block us
* until it finishes (and has updated the flat file copy of pg_database).
*
* Note that the lock is not held long, only until the end of this startup
* transaction. This is OK since we are already advertising our use of
* the database in the PGPROC array; anyone trying a DROP DATABASE after
* this point will see us there.
*
* Note: use of RowExclusiveLock here is reasonable because we envision
* our session as being a concurrent writer of the database. If we had a
* way of declaring a session as being guaranteed-read-only, we could use
* AccessShareLock for such sessions and thereby not conflict against
* CREATE DATABASE.
*/
if (!bootstrap)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
RowExclusiveLock);
/*
* Recheck the flat file copy of pg_database to make sure the target
* database hasn't gone away. If there was a concurrent DROP DATABASE,
* this ensures we will die cleanly without creating a mess.
*/
if (!bootstrap)
{
Oid dbid2;
Oid tsid2;
if (!FindMyDatabase(dbname, &dbid2, &tsid2) ||
dbid2 != MyDatabaseId || tsid2 != MyDatabaseTableSpace)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("It seems to have just been dropped or renamed.")));
}
/*
* Now we should be able to access the database directory safely. Verify
* it's there and looks reasonable.
*/
if (!bootstrap)
{
if (access(fullpath, F_OK) == -1)
{
if (errno == ENOENT)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("The database subdirectory \"%s\" is missing.",
fullpath)));
else
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not access directory \"%s\": %m",
fullpath)));
}
ValidatePgVersion(fullpath);
}
/*
* It's now possible to do real access to the system catalogs.
*
* Load relcache entries for the system catalogs. This must create at
* least the minimum set of "nailed-in" cache entries.
*/
RelationCacheInitializePhase2();
/*
* Figure out our postgres user id, and see if we are a superuser.
*
* In standalone mode and in the autovacuum process, we use a fixed id,
* otherwise we figure it out from the authenticated user name.
*/
if (bootstrap || autovacuum)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
if (!ThereIsAtLeastOneRole())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no roles are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" CREATEUSER;.",
username)));
}
else
{
/* normal multiuser case */
InitializeSessionUserId(username);
am_superuser = superuser();
}
/* set up ACL framework (so CheckMyDatabase can check permissions) */
initialize_acl();
/*
* Read the real pg_database row for our database, check permissions and
* set up database-specific GUC settings. We can't do this until all the
* database-access infrastructure is up. (Also, it wants to know if the
* user is a superuser, so the above stuff has to happen first.)
*/
if (!bootstrap)
CheckMyDatabase(dbname, am_superuser);
/*
* If we're trying to shut down, only superusers can connect.
*/
if (!am_superuser &&
MyProcPort != NULL &&
MyProcPort->canAcceptConnections == CAC_WAITBACKUP)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to connect during database shutdown")));
/*
* Check a normal user hasn't connected to a superuser reserved slot.
*/
if (!am_superuser &&
ReservedBackends > 0 &&
!HaveNFreeProcs(ReservedBackends))
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("connection limit exceeded for non-superusers")));
/*
* Initialize various default states that can't be set up until we've
* selected the active user and gotten the right GUC settings.
*/
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
return am_superuser;
}
/*
* lazy_vacuum_rel() -- perform LAZY VACUUM for one heap relation
*
* This routine vacuums a single heap, cleans out its indexes, and
* updates its relpages and reltuples statistics.
*
* At entry, we have already established a transaction and opened
* and locked the relation.
*
* The return value indicates whether this function has held off
* interrupts -- caller must RESUME_INTERRUPTS() after commit if true.
*/
bool
lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
BufferAccessStrategy bstrategy, List *updated_stats)
{
LVRelStats *vacrelstats;
Relation *Irel;
int nindexes;
BlockNumber possibly_freeable;
PGRUsage ru0;
TimestampTz starttime = 0;
bool heldoff = false;
pg_rusage_init(&ru0);
/* measure elapsed time iff autovacuum logging requires it */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration > 0)
starttime = GetCurrentTimestamp();
if (vacstmt->verbose)
elevel = INFO;
else
elevel = DEBUG2;
if (Gp_role == GP_ROLE_DISPATCH)
elevel = DEBUG2; /* vacuum and analyze messages aren't interesting from the QD */
#ifdef FAULT_INJECTOR
if (vacuumStatement_IsInAppendOnlyDropPhase(vacstmt))
{
FaultInjector_InjectFaultIfSet(
CompactionBeforeSegmentFileDropPhase,
DDLNotSpecified,
"", // databaseName
""); // tableName
}
if (vacummStatement_IsInAppendOnlyCleanupPhase(vacstmt))
{
FaultInjector_InjectFaultIfSet(
CompactionBeforeCleanupPhase,
DDLNotSpecified,
"", // databaseName
""); // tableName
}
#endif
/*
* MPP-23647. Update xid limits for heap as well as appendonly
* relations. This allows setting relfrozenxid to correct value
* for an appendonly (AO/CO) table.
*/
vac_strategy = bstrategy;
vacuum_set_xid_limits(vacstmt->freeze_min_age, onerel->rd_rel->relisshared,
&OldestXmin, &FreezeLimit);
/*
* Execute the various vacuum operations. Appendonly tables are treated
* differently.
*/
if (RelationIsAoRows(onerel) || RelationIsAoCols(onerel))
{
lazy_vacuum_aorel(onerel, vacstmt, updated_stats);
return false;
}
vacrelstats = (LVRelStats *) palloc0(sizeof(LVRelStats));
/* heap relation */
/* Set threshold for interesting free space = average request size */
/* XXX should we scale it up or down? Adjust vacuum.c too, if so */
vacrelstats->threshold = GetAvgFSMRequestSize(&onerel->rd_node);
vacrelstats->num_index_scans = 0;
/* Open all indexes of the relation */
vac_open_indexes(onerel, RowExclusiveLock, &nindexes, &Irel);
vacrelstats->hasindex = (nindexes > 0);
/* Do the vacuuming */
lazy_scan_heap(onerel, vacrelstats, Irel, nindexes, updated_stats);
/* Done with indexes */
vac_close_indexes(nindexes, Irel, NoLock);
/*
* Optionally truncate the relation.
*
* Don't even think about it unless we have a shot at releasing a goodly
* number of pages. Otherwise, the time taken isn't worth it.
*
* Note that after we've truncated the heap, it's too late to abort the
* transaction; doing so would lose the sinval messages needed to tell
* the other backends about the table being shrunk. We prevent interrupts
* in that case; caller is responsible for re-enabling them after
* committing the transaction.
*/
possibly_freeable = vacrelstats->rel_pages - vacrelstats->nonempty_pages;
if (possibly_freeable > 0 &&
(possibly_freeable >= REL_TRUNCATE_MINIMUM ||
possibly_freeable >= vacrelstats->rel_pages / REL_TRUNCATE_FRACTION))
{
HOLD_INTERRUPTS();
heldoff = true;
lazy_truncate_heap(onerel, vacrelstats);
}
/* Update shared free space map with final free space info */
lazy_update_fsm(onerel, vacrelstats);
if (vacrelstats->tot_free_pages > MaxFSMPages)
ereport(WARNING,
(errmsg("relation \"%s.%s\" contains more than \"max_fsm_pages\" pages with useful free space",
get_namespace_name(RelationGetNamespace(onerel)),
RelationGetRelationName(onerel)),
/* Only suggest VACUUM FULL if > 20% free */
(vacrelstats->tot_free_pages > vacrelstats->rel_pages * 0.20) ?
errhint("Consider using VACUUM FULL on this relation or increasing the configuration parameter \"max_fsm_pages\".") :
errhint("Consider increasing the configuration parameter \"max_fsm_pages\".")));
/* Update statistics in pg_class */
vac_update_relstats_from_list(onerel,
vacrelstats->rel_pages,
vacrelstats->rel_tuples,
vacrelstats->hasindex,
FreezeLimit,
updated_stats);
/* report results to the stats collector, too */
pgstat_report_vacuum(RelationGetRelid(onerel), onerel->rd_rel->relisshared,
true /*vacrelstats->scanned_all*/,
vacstmt->analyze, vacrelstats->rel_tuples);
if (gp_indexcheck_vacuum == INDEX_CHECK_ALL ||
(gp_indexcheck_vacuum == INDEX_CHECK_SYSTEM &&
PG_CATALOG_NAMESPACE == RelationGetNamespace(onerel)))
{
int i;
for (i = 0; i < nindexes; i++)
{
if (Irel[i]->rd_rel->relam == BTREE_AM_OID)
_bt_validate_vacuum(Irel[i], onerel, OldestXmin);
}
}
/* and log the action if appropriate */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0)
{
if (Log_autovacuum_min_duration == 0 ||
TimestampDifferenceExceeds(starttime, GetCurrentTimestamp(),
Log_autovacuum_min_duration))
ereport(LOG,
(errmsg("automatic vacuum of table \"%s.%s.%s\": index scans: %d\n"
"pages: %d removed, %d remain\n"
"tuples: %.0f removed, %.0f remain\n"
"system usage: %s",
get_database_name(MyDatabaseId),
get_namespace_name(RelationGetNamespace(onerel)),
RelationGetRelationName(onerel),
vacrelstats->num_index_scans,
vacrelstats->pages_removed, vacrelstats->rel_pages,
vacrelstats->tuples_deleted, vacrelstats->rel_tuples,
pg_rusage_show(&ru0))));
}
return heldoff;
}
/*
* Move tuples from pending pages into regular GIN structure.
*
* On first glance it looks completely not crash-safe. But if we crash
* after posting entries to the main index and before removing them from the
* pending list, it's okay because when we redo the posting later on, nothing
* bad will happen.
*
* fill_fsm indicates that ginInsertCleanup should add deleted pages
* to FSM otherwise caller is responsible to put deleted pages into
* FSM.
*
* If stats isn't null, we count deleted pending pages into the counts.
*/
void
ginInsertCleanup(GinState *ginstate, bool full_clean,
bool fill_fsm, IndexBulkDeleteResult *stats)
{
Relation index = ginstate->index;
Buffer metabuffer,
buffer;
Page metapage,
page;
GinMetaPageData *metadata;
MemoryContext opCtx,
oldCtx;
BuildAccumulator accum;
KeyArray datums;
BlockNumber blkno,
blknoFinish;
bool cleanupFinish = false;
bool fsm_vac = false;
Size workMemory;
bool inVacuum = (stats == NULL);
/*
* We would like to prevent concurrent cleanup process. For that we will
* lock metapage in exclusive mode using LockPage() call. Nobody other
* will use that lock for metapage, so we keep possibility of concurrent
* insertion into pending list
*/
if (inVacuum)
{
/*
* We are called from [auto]vacuum/analyze or gin_clean_pending_list()
* and we would like to wait concurrent cleanup to finish.
*/
LockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
workMemory =
(IsAutoVacuumWorkerProcess() && autovacuum_work_mem != -1) ?
autovacuum_work_mem : maintenance_work_mem;
}
else
{
/*
* We are called from regular insert and if we see concurrent cleanup
* just exit in hope that concurrent process will clean up pending
* list.
*/
if (!ConditionalLockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock))
return;
workMemory = work_mem;
}
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
LockBuffer(metabuffer, GIN_SHARE);
metapage = BufferGetPage(metabuffer);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/* Nothing to do */
UnlockReleaseBuffer(metabuffer);
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
return;
}
/*
* Remember a tail page to prevent infinite cleanup if other backends add
* new tuples faster than we can cleanup.
*/
blknoFinish = metadata->tail;
/*
* Read and lock head of pending list
*/
blkno = metadata->head;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* Initialize. All temporary space will be in opCtx
*/
opCtx = AllocSetContextCreate(CurrentMemoryContext,
"GIN insert cleanup temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCtx = MemoryContextSwitchTo(opCtx);
initKeyArray(&datums, 128);
ginInitBA(&accum);
accum.ginstate = ginstate;
/*
* At the top of this loop, we have pin and lock on the current page of
* the pending list. However, we'll release that before exiting the loop.
* Note we also have pin but not lock on the metapage.
*/
for (;;)
{
Assert(!GinPageIsDeleted(page));
/*
* Are we walk through the page which as we remember was a tail when
* we start our cleanup? But if caller asks us to clean up whole
* pending list then ignore old tail, we will work until list becomes
* empty.
*/
if (blkno == blknoFinish && full_clean == false)
cleanupFinish = true;
/*
* read page's datums into accum
*/
processPendingPage(&accum, &datums, page, FirstOffsetNumber);
vacuum_delay_point();
/*
* Is it time to flush memory to disk? Flush if we are at the end of
* the pending list, or if we have a full row and memory is getting
* full.
*/
if (GinPageGetOpaque(page)->rightlink == InvalidBlockNumber ||
(GinPageHasFullRow(page) &&
(accum.allocatedMemory >= workMemory * 1024L)))
{
ItemPointerData *list;
uint32 nlist;
Datum key;
GinNullCategory category;
OffsetNumber maxoff,
attnum;
/*
* Unlock current page to increase performance. Changes of page
* will be checked later by comparing maxoff after completion of
* memory flush.
*/
maxoff = PageGetMaxOffsetNumber(page);
LockBuffer(buffer, GIN_UNLOCK);
/*
* Moving collected data into regular structure can take
* significant amount of time - so, run it without locking pending
* list.
*/
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
{
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
vacuum_delay_point();
}
/*
* Lock the whole list to remove pages
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsDeleted(page));
/*
* While we left the page unlocked, more stuff might have gotten
* added to it. If so, process those entries immediately. There
* shouldn't be very many, so we don't worry about the fact that
* we're doing this with exclusive lock. Insertion algorithm
* guarantees that inserted row(s) will not continue on next page.
* NOTE: intentionally no vacuum_delay_point in this loop.
*/
if (PageGetMaxOffsetNumber(page) != maxoff)
{
ginInitBA(&accum);
processPendingPage(&accum, &datums, page, maxoff + 1);
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
}
/*
* Remember next page - it will become the new list head
*/
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer); /* shiftList will do exclusive
* locking */
/*
* remove read pages from pending list, at this point all content
* of read pages is in regular structure
*/
shiftList(index, metabuffer, blkno, fill_fsm, stats);
/* At this point, some pending pages have been freed up */
fsm_vac = true;
Assert(blkno == metadata->head);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* if we removed the whole pending list or we cleanup tail (which
* we remembered on start our cleanup process) then just exit
*/
if (blkno == InvalidBlockNumber || cleanupFinish)
break;
/*
* release memory used so far and reinit state
*/
MemoryContextReset(opCtx);
initKeyArray(&datums, datums.maxvalues);
ginInitBA(&accum);
}
else
{
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer);
}
/*
* Read next page in pending list
*/
vacuum_delay_point();
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
}
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
ReleaseBuffer(metabuffer);
/*
* As pending list pages can have a high churn rate, it is desirable to
* recycle them immediately to the FreeSpace Map when ordinary backends
* clean the list.
*/
if (fsm_vac && fill_fsm)
IndexFreeSpaceMapVacuum(index);
/* Clean up temporary space */
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(opCtx);
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* The database can be specified by name, using the in_dbname parameter, or by
* OID, using the dboid parameter. In the latter case, the actual database
* name can be returned to the caller in out_dbname. If out_dbname isn't
* NULL, it must point to a buffer of size NAMEDATALEN.
*
* In bootstrap mode no parameters are used. The autovacuum launcher process
* doesn't use any parameters either, because it only goes far enough to be
* able to read pg_database; it doesn't connect to any particular database.
* In walsender mode only username is used.
*
* As of PostgreSQL 8.2, we expect InitProcess() was already called, so we
* already have a PGPROC struct ... but it's not completely filled in yet.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
void
InitPostgres(const char *in_dbname, Oid dboid, const char *username,
char *out_dbname)
{
bool bootstrap = IsBootstrapProcessingMode();
bool am_superuser;
char *fullpath;
char dbname[NAMEDATALEN];
elog(DEBUG3, "InitPostgres");
/*
* Add my PGPROC struct to the ProcArray.
*
* Once I have done this, I am visible to other backends!
*/
InitProcessPhase2();
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data.
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
SharedInvalBackendInit(false);
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend ID: %d", MyBackendId);
/* Now that we have a BackendId, we can participate in ProcSignal */
ProcSignalInit(MyBackendId);
/*
* Also set up timeout handlers needed for backend operation. We need
* these in every case except bootstrap.
*/
if (!bootstrap)
{
RegisterTimeout(DEADLOCK_TIMEOUT, CheckDeadLock);
RegisterTimeout(STATEMENT_TIMEOUT, StatementTimeoutHandler);
RegisterTimeout(LOCK_TIMEOUT, LockTimeoutHandler);
}
/*
* bufmgr needs another initialization call too
*/
InitBufferPoolBackend();
/*
* Initialize local process's access to XLOG.
*/
if (IsUnderPostmaster)
{
/*
* The postmaster already started the XLOG machinery, but we need to
* call InitXLOGAccess(), if the system isn't in hot-standby mode.
* This is handled by calling RecoveryInProgress and ignoring the
* result.
*/
(void) RecoveryInProgress();
}
else
{
/*
* We are either a bootstrap process or a standalone backend. Either
* way, start up the XLOG machinery, and register to have it closed
* down at exit.
*/
StartupXLOG();
on_shmem_exit(ShutdownXLOG, 0);
}
/*
* Initialize the relation cache and the system catalog caches. Note that
* no catalog access happens here; we only set up the hashtable structure.
* We must do this before starting a transaction because transaction abort
* would try to touch these hashtables.
*/
RelationCacheInitialize();
InitCatalogCache();
InitPlanCache();
/* Initialize portal manager */
EnablePortalManager();
/* Initialize stats collection --- must happen before first xact */
if (!bootstrap)
pgstat_initialize();
/*
* Load relcache entries for the shared system catalogs. This must create
* at least entries for pg_database and catalogs used for authentication.
*/
RelationCacheInitializePhase2();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This is the
* first before_shmem_exit callback we register; thus, this will be the
* last thing we do before low-level modules like the buffer manager begin
* to close down. We need to have this in place before we begin our first
* transaction --- if we fail during the initialization transaction, as is
* entirely possible, we need the AbortTransaction call to clean up.
*/
before_shmem_exit(ShutdownPostgres, 0);
/* The autovacuum launcher is done here */
if (IsAutoVacuumLauncherProcess())
return;
/*
* Start a new transaction here before first access to db, and get a
* snapshot. We don't have a use for the snapshot itself, but we're
* interested in the secondary effect that it sets RecentGlobalXmin. (This
* is critical for anything that reads heap pages, because HOT may decide
* to prune them even if the process doesn't attempt to modify any
* tuples.)
*/
if (!bootstrap)
{
/* statement_timestamp must be set for timeouts to work correctly */
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
/*
* transaction_isolation will have been set to the default by the
* above. If the default is "serializable", and we are in hot
* standby, we will fail if we don't change it to something lower.
* Fortunately, "read committed" is plenty good enough.
*/
XactIsoLevel = XACT_READ_COMMITTED;
(void) GetTransactionSnapshot();
}
/*
* Perform client authentication if necessary, then figure out our
* postgres user ID, and see if we are a superuser.
*
* In standalone mode and in autovacuum worker processes, we use a fixed
* ID, otherwise we figure it out from the authenticated user name.
*/
if (bootstrap || IsAutoVacuumWorkerProcess())
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
if (!ThereIsAtLeastOneRole())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no roles are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" SUPERUSER;.",
username)));
}
else if (IsBackgroundWorker)
{
if (username == NULL)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else
{
InitializeSessionUserId(username);
am_superuser = superuser();
}
}
else
{
/* normal multiuser case */
Assert(MyProcPort != NULL);
PerformAuthentication(MyProcPort);
InitializeSessionUserId(username);
am_superuser = superuser();
}
/*
* If we're trying to shut down, only superusers can connect, and new
* replication connections are not allowed.
*/
if ((!am_superuser || am_walsender) &&
MyProcPort != NULL &&
MyProcPort->canAcceptConnections == CAC_WAITBACKUP)
{
if (am_walsender)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("new replication connections are not allowed during database shutdown")));
else
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to connect during database shutdown")));
}
/*
* Binary upgrades only allowed super-user connections
*/
if (IsBinaryUpgrade && !am_superuser)
{
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to connect in binary upgrade mode")));
}
/*
* The last few connections slots are reserved for superusers. Although
* replication connections currently require superuser privileges, we
* don't allow them to consume the reserved slots, which are intended for
* interactive use.
*/
if ((!am_superuser || am_walsender) &&
ReservedBackends > 0 &&
!HaveNFreeProcs(ReservedBackends))
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("remaining connection slots are reserved for non-replication superuser connections")));
/* Check replication permissions needed for walsender processes. */
if (am_walsender)
{
Assert(!bootstrap);
if (!superuser() && !has_rolreplication(GetUserId()))
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser or replication role to start walsender")));
}
/*
* If this is a plain walsender only supporting physical replication, we
* don't want to connect to any particular database. Just finish the
* backend startup by processing any options from the startup packet, and
* we're done.
*/
if (am_walsender && !am_db_walsender)
{
/* process any options passed in the startup packet */
if (MyProcPort != NULL)
process_startup_options(MyProcPort, am_superuser);
/* Apply PostAuthDelay as soon as we've read all options */
if (PostAuthDelay > 0)
pg_usleep(PostAuthDelay * 1000000L);
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
pgstat_bestart();
/* close the transaction we started above */
CommitTransactionCommand();
return;
}
/*
* Set up the global variables holding database id and default tablespace.
* But note we won't actually try to touch the database just yet.
*
* We take a shortcut in the bootstrap case, otherwise we have to look up
* the db's entry in pg_database.
*/
if (bootstrap)
{
MyDatabaseId = TemplateDbOid;
MyDatabaseTableSpace = DEFAULTTABLESPACE_OID;
}
else if (in_dbname != NULL)
{
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTuple(in_dbname);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", in_dbname)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
/* take database name from the caller, just for paranoia */
strlcpy(dbname, in_dbname, sizeof(dbname));
}
else if (OidIsValid(dboid))
{
/* caller specified database by OID */
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTupleByOid(dboid);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database %u does not exist", dboid)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
Assert(MyDatabaseId == dboid);
strlcpy(dbname, NameStr(dbform->datname), sizeof(dbname));
/* pass the database name back to the caller */
if (out_dbname)
strcpy(out_dbname, dbname);
}
else
{
/*
* If this is a background worker not bound to any particular
* database, we're done now. Everything that follows only makes
* sense if we are bound to a specific database. We do need to
* close the transaction we started before returning.
*/
if (!bootstrap)
CommitTransactionCommand();
return;
}
/*
* Now, take a writer's lock on the database we are trying to connect to.
* If there is a concurrently running DROP DATABASE on that database, this
* will block us until it finishes (and has committed its update of
* pg_database).
*
* Note that the lock is not held long, only until the end of this startup
* transaction. This is OK since we will advertise our use of the
* database in the ProcArray before dropping the lock (in fact, that's the
* next thing to do). Anyone trying a DROP DATABASE after this point will
* see us in the array once they have the lock. Ordering is important for
* this because we don't want to advertise ourselves as being in this
* database until we have the lock; otherwise we create what amounts to a
* deadlock with CountOtherDBBackends().
*
* Note: use of RowExclusiveLock here is reasonable because we envision
* our session as being a concurrent writer of the database. If we had a
* way of declaring a session as being guaranteed-read-only, we could use
* AccessShareLock for such sessions and thereby not conflict against
* CREATE DATABASE.
*/
if (!bootstrap)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
RowExclusiveLock);
/*
* Now we can mark our PGPROC entry with the database ID.
*
* We assume this is an atomic store so no lock is needed; though actually
* things would work fine even if it weren't atomic. Anyone searching the
* ProcArray for this database's ID should hold the database lock, so they
* would not be executing concurrently with this store. A process looking
* for another database's ID could in theory see a chance match if it read
* a partially-updated databaseId value; but as long as all such searches
* wait and retry, as in CountOtherDBBackends(), they will certainly see
* the correct value on their next try.
*/
MyProc->databaseId = MyDatabaseId;
/*
* We established a catalog snapshot while reading pg_authid and/or
* pg_database; but until we have set up MyDatabaseId, we won't react to
* incoming sinval messages for unshared catalogs, so we won't realize it
* if the snapshot has been invalidated. Assume it's no good anymore.
*/
InvalidateCatalogSnapshot();
/*
* Recheck pg_database to make sure the target database hasn't gone away.
* If there was a concurrent DROP DATABASE, this ensures we will die
* cleanly without creating a mess.
*/
if (!bootstrap)
{
HeapTuple tuple;
tuple = GetDatabaseTuple(dbname);
if (!HeapTupleIsValid(tuple) ||
MyDatabaseId != HeapTupleGetOid(tuple) ||
MyDatabaseTableSpace != ((Form_pg_database) GETSTRUCT(tuple))->dattablespace)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", dbname),
errdetail("It seems to have just been dropped or renamed.")));
}
/*
* Now we should be able to access the database directory safely. Verify
* it's there and looks reasonable.
*/
fullpath = GetDatabasePath(MyDatabaseId, MyDatabaseTableSpace);
if (!bootstrap)
{
if (access(fullpath, F_OK) == -1)
{
if (errno == ENOENT)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("The database subdirectory \"%s\" is missing.",
fullpath)));
else
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not access directory \"%s\": %m",
fullpath)));
}
ValidatePgVersion(fullpath);
}
SetDatabasePath(fullpath);
/*
* It's now possible to do real access to the system catalogs.
*
* Load relcache entries for the system catalogs. This must create at
* least the minimum set of "nailed-in" cache entries.
*/
RelationCacheInitializePhase3();
/* set up ACL framework (so CheckMyDatabase can check permissions) */
initialize_acl();
/*
* Re-read the pg_database row for our database, check permissions and set
* up database-specific GUC settings. We can't do this until all the
* database-access infrastructure is up. (Also, it wants to know if the
* user is a superuser, so the above stuff has to happen first.)
*/
if (!bootstrap)
CheckMyDatabase(dbname, am_superuser);
/*
* Now process any command-line switches and any additional GUC variable
* settings passed in the startup packet. We couldn't do this before
* because we didn't know if client is a superuser.
*/
if (MyProcPort != NULL)
process_startup_options(MyProcPort, am_superuser);
/* Process pg_db_role_setting options */
process_settings(MyDatabaseId, GetSessionUserId());
/* Apply PostAuthDelay as soon as we've read all options */
if (PostAuthDelay > 0)
pg_usleep(PostAuthDelay * 1000000L);
/*
* Initialize various default states that can't be set up until we've
* selected the active user and gotten the right GUC settings.
*/
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
}
/*
* lazy_vacuum_rel() -- perform LAZY VACUUM for one heap relation
*
* This routine vacuums a single heap, cleans out its indexes, and
* updates its relpages and reltuples statistics.
*
* At entry, we have already established a transaction and opened
* and locked the relation.
*/
void
lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
BufferAccessStrategy bstrategy)
{
LVRelStats *vacrelstats;
Relation *Irel;
int nindexes;
BlockNumber possibly_freeable;
PGRUsage ru0;
TimestampTz starttime = 0;
long secs;
int usecs;
double read_rate,
write_rate;
bool scan_all;
TransactionId freezeTableLimit;
BlockNumber new_rel_pages;
double new_rel_tuples;
BlockNumber new_rel_allvisible;
TransactionId new_frozen_xid;
/* measure elapsed time iff autovacuum logging requires it */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0)
{
pg_rusage_init(&ru0);
starttime = GetCurrentTimestamp();
}
if (vacstmt->options & VACOPT_VERBOSE)
elevel = INFO;
else
elevel = DEBUG2;
vac_strategy = bstrategy;
vacuum_set_xid_limits(vacstmt->freeze_min_age, vacstmt->freeze_table_age,
onerel->rd_rel->relisshared,
&OldestXmin, &FreezeLimit, &freezeTableLimit);
scan_all = TransactionIdPrecedesOrEquals(onerel->rd_rel->relfrozenxid,
freezeTableLimit);
vacrelstats = (LVRelStats *) palloc0(sizeof(LVRelStats));
vacrelstats->old_rel_pages = onerel->rd_rel->relpages;
vacrelstats->old_rel_tuples = onerel->rd_rel->reltuples;
vacrelstats->num_index_scans = 0;
/* Open all indexes of the relation */
vac_open_indexes(onerel, RowExclusiveLock, &nindexes, &Irel);
vacrelstats->hasindex = (nindexes > 0);
/* Do the vacuuming */
lazy_scan_heap(onerel, vacrelstats, Irel, nindexes, scan_all);
/* Done with indexes */
vac_close_indexes(nindexes, Irel, NoLock);
/*
* Optionally truncate the relation.
*
* Don't even think about it unless we have a shot at releasing a goodly
* number of pages. Otherwise, the time taken isn't worth it.
*/
possibly_freeable = vacrelstats->rel_pages - vacrelstats->nonempty_pages;
if (possibly_freeable > 0 &&
(possibly_freeable >= REL_TRUNCATE_MINIMUM ||
possibly_freeable >= vacrelstats->rel_pages / REL_TRUNCATE_FRACTION))
lazy_truncate_heap(onerel, vacrelstats);
/* Vacuum the Free Space Map */
FreeSpaceMapVacuum(onerel);
/*
* Update statistics in pg_class.
*
* A corner case here is that if we scanned no pages at all because every
* page is all-visible, we should not update relpages/reltuples, because
* we have no new information to contribute. In particular this keeps
* us from replacing relpages=reltuples=0 (which means "unknown tuple
* density") with nonzero relpages and reltuples=0 (which means "zero
* tuple density") unless there's some actual evidence for the latter.
*
* We do update relallvisible even in the corner case, since if the
* table is all-visible we'd definitely like to know that. But clamp
* the value to be not more than what we're setting relpages to.
*
* Also, don't change relfrozenxid if we skipped any pages, since then
* we don't know for certain that all tuples have a newer xmin.
*/
new_rel_pages = vacrelstats->rel_pages;
new_rel_tuples = vacrelstats->new_rel_tuples;
if (vacrelstats->scanned_pages == 0 && new_rel_pages > 0)
{
new_rel_pages = vacrelstats->old_rel_pages;
new_rel_tuples = vacrelstats->old_rel_tuples;
}
new_rel_allvisible = visibilitymap_count(onerel);
if (new_rel_allvisible > new_rel_pages)
new_rel_allvisible = new_rel_pages;
new_frozen_xid = FreezeLimit;
if (vacrelstats->scanned_pages < vacrelstats->rel_pages)
new_frozen_xid = InvalidTransactionId;
vac_update_relstats(onerel,
new_rel_pages,
new_rel_tuples,
new_rel_allvisible,
vacrelstats->hasindex,
new_frozen_xid);
/* report results to the stats collector, too */
pgstat_report_vacuum(RelationGetRelid(onerel),
onerel->rd_rel->relisshared,
new_rel_tuples);
/* and log the action if appropriate */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0)
{
TimestampTz endtime = GetCurrentTimestamp();
if (Log_autovacuum_min_duration == 0 ||
TimestampDifferenceExceeds(starttime, endtime,
Log_autovacuum_min_duration))
{
TimestampDifference(starttime, endtime, &secs, &usecs);
read_rate = 0;
write_rate = 0;
if ((secs > 0) || (usecs > 0))
{
read_rate = (double) BLCKSZ * VacuumPageMiss / (1024 * 1024) /
(secs + usecs / 1000000.0);
write_rate = (double) BLCKSZ * VacuumPageDirty / (1024 * 1024) /
(secs + usecs / 1000000.0);
}
ereport(LOG,
(errmsg("automatic vacuum of table \"%s.%s.%s\": index scans: %d\n"
"pages: %d removed, %d remain\n"
"tuples: %.0f removed, %.0f remain\n"
"buffer usage: %d hits, %d misses, %d dirtied\n"
"avg read rate: %.3f MiB/s, avg write rate: %.3f MiB/s\n"
"system usage: %s",
get_database_name(MyDatabaseId),
get_namespace_name(RelationGetNamespace(onerel)),
RelationGetRelationName(onerel),
vacrelstats->num_index_scans,
vacrelstats->pages_removed,
vacrelstats->rel_pages,
vacrelstats->tuples_deleted,
vacrelstats->new_rel_tuples,
VacuumPageHit,
VacuumPageMiss,
VacuumPageDirty,
read_rate,write_rate,
pg_rusage_show(&ru0))));
}
}
}
/*
* 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
DtmXactCallback(XactEvent event, void *arg)
{
//XTM_INFO("%d: DtmXactCallbackevent=%d nextxid=%d\n", getpid(), event, DtmNextXid);
switch (event)
{
case XACT_EVENT_START:
//XTM_INFO("%d: normal=%d, initialized=%d, replication=%d, bgw=%d, vacuum=%d\n",
// getpid(), IsNormalProcessingMode(), dtm->initialized, MMDoReplication, IsBackgroundWorker, IsAutoVacuumWorkerProcess());
if (IsNormalProcessingMode() && dtm->initialized && MMDoReplication && !am_walsender && !IsBackgroundWorker && !IsAutoVacuumWorkerProcess()) {
MMBeginTransaction();
}
break;
#if 0
case XACT_EVENT_PRE_COMMIT:
case XACT_EVENT_PARALLEL_PRE_COMMIT:
{
TransactionId xid = GetCurrentTransactionIdIfAny();
if (!MMIsDistributedTrans && TransactionIdIsValid(xid)) {
XTM_INFO("%d: Will ignore transaction %u\n", getpid(), xid);
MMMarkTransAsLocal(xid);
}
break;
}
#endif
case XACT_EVENT_COMMIT:
case XACT_EVENT_ABORT:
if (TransactionIdIsValid(DtmNextXid))
{
if (!DtmVoted) {
ArbiterSetTransStatus(DtmNextXid, TRANSACTION_STATUS_ABORTED, false);
}
if (event == XACT_EVENT_COMMIT)
{
/*
* Now transaction status is already written in CLOG,
* so we can remove information about it from hash table
*/
LWLockAcquire(dtm->hashLock, LW_EXCLUSIVE);
hash_search(xid_in_doubt, &DtmNextXid, HASH_REMOVE, NULL);
LWLockRelease(dtm->hashLock);
}
#if 0 /* should be handled now using DtmVoted flag */
else
{
/*
* Transaction at the node can be aborted because of transaction failure at some other node
* before it starts doing anything and assigned Xid, in this case Postgres is not calling SetTransactionStatus,
* so we have to send report to DTMD here
*/
if (!TransactionIdIsValid(GetCurrentTransactionIdIfAny())) {
XTM_INFO("%d: abort transation on DTMD\n", getpid());
ArbiterSetTransStatus(DtmNextXid, TRANSACTION_STATUS_ABORTED, false);
}
}
#endif
DtmNextXid = InvalidTransactionId;
DtmLastSnapshot = NULL;
}
MMIsDistributedTrans = false;
break;
default:
break;
}
}
/*
* 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();
}
}
}
}
/*
* InitProcess -- initialize a per-process data structure for this backend
*/
void
InitProcess(void)
{
/* use volatile pointer to prevent code rearrangement */
volatile PROC_HDR *procglobal = ProcGlobal;
/*
* ProcGlobal should be set up already (if we are a backend, we inherit
* this by fork() or EXEC_BACKEND mechanism from the postmaster).
*/
if (procglobal == NULL)
elog(PANIC, "proc header uninitialized");
if (MyProc != NULL)
elog(ERROR, "you already exist");
/*
* Try to get a proc struct from the free list. If this fails, we must be
* out of PGPROC structures (not to mention semaphores).
*
* While we are holding the ProcStructLock, also copy the current shared
* estimate of spins_per_delay to local storage.
*/
SpinLockAcquire(ProcStructLock);
set_spins_per_delay(procglobal->spins_per_delay);
if (IsAnyAutoVacuumProcess())
MyProc = procglobal->autovacFreeProcs;
else if (IsBackgroundWorker)
MyProc = procglobal->bgworkerFreeProcs;
else
MyProc = procglobal->freeProcs;
if (MyProc != NULL)
{
if (IsAnyAutoVacuumProcess())
procglobal->autovacFreeProcs = (PGPROC *) MyProc->links.next;
else if (IsBackgroundWorker)
procglobal->bgworkerFreeProcs = (PGPROC *) MyProc->links.next;
else
procglobal->freeProcs = (PGPROC *) MyProc->links.next;
SpinLockRelease(ProcStructLock);
}
else
{
/*
* If we reach here, all the PGPROCs are in use. This is one of the
* possible places to detect "too many backends", so give the standard
* error message. XXX do we need to give a different failure message
* in the autovacuum case?
*/
SpinLockRelease(ProcStructLock);
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("sorry, too many clients already")));
}
MyPgXact = &ProcGlobal->allPgXact[MyProc->pgprocno];
/*
* Now that we have a PGPROC, mark ourselves as an active postmaster
* child; this is so that the postmaster can detect it if we exit without
* cleaning up. (XXX autovac launcher currently doesn't participate in
* this; it probably should.)
*/
if (IsUnderPostmaster && !IsAutoVacuumLauncherProcess())
MarkPostmasterChildActive();
/*
* Initialize all fields of MyProc, except for those previously
* initialized by InitProcGlobal.
*/
SHMQueueElemInit(&(MyProc->links));
MyProc->waitStatus = STATUS_OK;
MyProc->lxid = InvalidLocalTransactionId;
MyProc->fpVXIDLock = false;
MyProc->fpLocalTransactionId = InvalidLocalTransactionId;
MyPgXact->xid = InvalidTransactionId;
MyPgXact->xmin = InvalidTransactionId;
MyProc->pid = MyProcPid;
/* backendId, databaseId and roleId will be filled in later */
MyProc->backendId = InvalidBackendId;
MyProc->databaseId = InvalidOid;
MyProc->roleId = InvalidOid;
MyPgXact->delayChkpt = false;
MyPgXact->vacuumFlags = 0;
/* NB -- autovac launcher intentionally does not set IS_AUTOVACUUM */
if (IsAutoVacuumWorkerProcess())
MyPgXact->vacuumFlags |= PROC_IS_AUTOVACUUM;
MyProc->lwWaiting = false;
MyProc->lwWaitMode = 0;
MyProc->waitLock = NULL;
MyProc->waitProcLock = NULL;
#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
MyProc->recoveryConflictPending = false;
/* Initialize fields for sync rep */
MyProc->waitLSN = 0;
MyProc->syncRepState = SYNC_REP_NOT_WAITING;
SHMQueueElemInit(&(MyProc->syncRepLinks));
/*
* Acquire ownership of the PGPROC's latch, so that we can use WaitLatch
* on it. That allows us to repoint the process latch, which so far
* points to process local one, to the shared one.
*/
OwnLatch(&MyProc->procLatch);
SwitchToSharedLatch();
/*
* We might be reusing a semaphore that belonged to a failed process. So
* be careful and reinitialize its value here. (This is not strictly
* necessary anymore, but seems like a good idea for cleanliness.)
*/
PGSemaphoreReset(&MyProc->sem);
/*
* Arrange to clean up at backend exit.
*/
on_shmem_exit(ProcKill, 0);
/*
* Now that we have a PGPROC, we could try to acquire locks, so initialize
* local state needed for LWLocks, and the deadlock checker.
*/
InitLWLockAccess();
InitDeadLockChecking();
}
Datum
ginvacuumcleanup(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
Relation index = info->index;
bool needLock;
BlockNumber npages,
blkno;
BlockNumber totFreePages;
GinState ginstate;
GinStatsData idxStat;
/*
* In an autovacuum analyze, we want to clean up pending insertions.
* Otherwise, an ANALYZE-only call is a no-op.
*/
if (info->analyze_only)
{
if (IsAutoVacuumWorkerProcess())
{
initGinState(&ginstate, index);
ginInsertCleanup(&ginstate, true, stats);
}
PG_RETURN_POINTER(stats);
}
/*
* Set up all-zero stats and cleanup pending inserts if ginbulkdelete
* wasn't called
*/
if (stats == NULL)
{
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
initGinState(&ginstate, index);
ginInsertCleanup(&ginstate, true, stats);
}
memset(&idxStat, 0, sizeof(idxStat));
/*
* XXX we always report the heap tuple count as the number of index
* entries. This is bogus if the index is partial, but it's real hard to
* tell how many distinct heap entries are referenced by a GIN index.
*/
stats->num_index_tuples = info->num_heap_tuples;
stats->estimated_count = info->estimated_count;
/*
* Need lock unless it's local to this backend.
*/
needLock = !RELATION_IS_LOCAL(index);
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
npages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
totFreePages = 0;
for (blkno = GIN_ROOT_BLKNO; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, GIN_SHARE);
page = (Page) BufferGetPage(buffer);
if (GinPageIsDeleted(page))
{
Assert(blkno != GIN_ROOT_BLKNO);
RecordFreeIndexPage(index, blkno);
totFreePages++;
}
else if (GinPageIsData(page))
{
idxStat.nDataPages++;
}
else if (!GinPageIsList(page))
{
idxStat.nEntryPages++;
if (GinPageIsLeaf(page))
idxStat.nEntries += PageGetMaxOffsetNumber(page);
}
UnlockReleaseBuffer(buffer);
}
/* Update the metapage with accurate page and entry counts */
idxStat.nTotalPages = npages;
ginUpdateStats(info->index, &idxStat);
/* Finally, vacuum the FSM */
IndexFreeSpaceMapVacuum(info->index);
stats->pages_free = totFreePages;
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
stats->num_pages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
PG_RETURN_POINTER(stats);
}
IndexBulkDeleteResult *
ginbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation index = info->index;
BlockNumber blkno = GIN_ROOT_BLKNO;
GinVacuumState gvs;
Buffer buffer;
BlockNumber rootOfPostingTree[BLCKSZ / (sizeof(IndexTupleData) + sizeof(ItemId))];
uint32 nRoot;
gvs.tmpCxt = AllocSetContextCreate(CurrentMemoryContext,
"Gin vacuum temporary context",
ALLOCSET_DEFAULT_SIZES);
gvs.index = index;
gvs.callback = callback;
gvs.callback_state = callback_state;
gvs.strategy = info->strategy;
initGinState(&gvs.ginstate, index);
/* first time through? */
if (stats == NULL)
{
/* Yes, so initialize stats to zeroes */
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/*
* and cleanup any pending inserts
*/
ginInsertCleanup(&gvs.ginstate, !IsAutoVacuumWorkerProcess(),
false, stats);
}
/* we'll re-count the tuples each time */
stats->num_index_tuples = 0;
gvs.result = stats;
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
/* find leaf page */
for (;;)
{
Page page = BufferGetPage(buffer);
IndexTuple itup;
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsData(page));
if (GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
LockBuffer(buffer, GIN_EXCLUSIVE);
if (blkno == GIN_ROOT_BLKNO && !GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
continue; /* check it one more */
}
break;
}
Assert(PageGetMaxOffsetNumber(page) >= FirstOffsetNumber);
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, FirstOffsetNumber));
blkno = GinGetDownlink(itup);
Assert(blkno != InvalidBlockNumber);
UnlockReleaseBuffer(buffer);
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
}
/* right now we found leftmost page in entry's BTree */
for (;;)
{
Page page = BufferGetPage(buffer);
Page resPage;
uint32 i;
Assert(!GinPageIsData(page));
resPage = ginVacuumEntryPage(&gvs, buffer, rootOfPostingTree, &nRoot);
blkno = GinPageGetOpaque(page)->rightlink;
if (resPage)
{
START_CRIT_SECTION();
PageRestoreTempPage(resPage, page);
MarkBufferDirty(buffer);
xlogVacuumPage(gvs.index, buffer);
UnlockReleaseBuffer(buffer);
END_CRIT_SECTION();
}
else
{
UnlockReleaseBuffer(buffer);
}
vacuum_delay_point();
for (i = 0; i < nRoot; i++)
{
ginVacuumPostingTree(&gvs, rootOfPostingTree[i]);
vacuum_delay_point();
}
if (blkno == InvalidBlockNumber) /* rightmost page */
break;
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, GIN_EXCLUSIVE);
}
MemoryContextDelete(gvs.tmpCxt);
return gvs.result;
}
/*
* CheckMyDatabase -- fetch information from the pg_database entry for our DB
*/
static void
CheckMyDatabase(const char *name, bool am_superuser)
{
HeapTuple tup;
Form_pg_database dbform;
char *collate;
char *ctype;
/* Fetch our pg_database row normally, via syscache */
tup = SearchSysCache1(DATABASEOID, ObjectIdGetDatum(MyDatabaseId));
if (!HeapTupleIsValid(tup))
elog(ERROR, "cache lookup failed for database %u", MyDatabaseId);
dbform = (Form_pg_database) GETSTRUCT(tup);
/* This recheck is strictly paranoia */
if (strcmp(name, NameStr(dbform->datname)) != 0)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" has disappeared from pg_database",
name),
errdetail("Database OID %u now seems to belong to \"%s\".",
MyDatabaseId, NameStr(dbform->datname))));
/*
* Check permissions to connect to the database.
*
* These checks are not enforced when in standalone mode, so that there is
* a way to recover from disabling all access to all databases, for
* example "UPDATE pg_database SET datallowconn = false;".
*
* We do not enforce them for autovacuum worker processes either.
*/
if (IsUnderPostmaster && !IsAutoVacuumWorkerProcess())
{
/*
* Check that the database is currently allowing connections.
*/
if (!dbform->datallowconn)
ereport(FATAL,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("database \"%s\" is not currently accepting connections",
name)));
/*
* Check privilege to connect to the database. (The am_superuser test
* is redundant, but since we have the flag, might as well check it
* and save a few cycles.)
*/
if (!am_superuser &&
pg_database_aclcheck(MyDatabaseId, GetUserId(),
ACL_CONNECT) != ACLCHECK_OK)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied for database \"%s\"", name),
errdetail("User does not have CONNECT privilege.")));
/*
* Check connection limit for this database.
*
* There is a race condition here --- we create our PGPROC before
* checking for other PGPROCs. If two backends did this at about the
* same time, they might both think they were over the limit, while
* ideally one should succeed and one fail. Getting that to work
* exactly seems more trouble than it is worth, however; instead we
* just document that the connection limit is approximate.
*/
if (dbform->datconnlimit >= 0 &&
!am_superuser &&
CountDBBackends(MyDatabaseId) > dbform->datconnlimit)
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("too many connections for database \"%s\"",
name)));
}
/*
* OK, we're golden. Next to-do item is to save the encoding info out of
* the pg_database tuple.
*/
SetDatabaseEncoding(dbform->encoding);
/* Record it as a GUC internal option, too */
SetConfigOption("server_encoding", GetDatabaseEncodingName(),
PGC_INTERNAL, PGC_S_OVERRIDE);
/* If we have no other source of client_encoding, use server encoding */
SetConfigOption("client_encoding", GetDatabaseEncodingName(),
PGC_BACKEND, PGC_S_DYNAMIC_DEFAULT);
/* assign locale variables */
collate = NameStr(dbform->datcollate);
ctype = NameStr(dbform->datctype);
if (pg_perm_setlocale(LC_COLLATE, collate) == NULL)
ereport(FATAL,
(errmsg("database locale is incompatible with operating system"),
errdetail("The database was initialized with LC_COLLATE \"%s\", "
" which is not recognized by setlocale().", collate),
errhint("Recreate the database with another locale or install the missing locale.")));
if (pg_perm_setlocale(LC_CTYPE, ctype) == NULL)
ereport(FATAL,
(errmsg("database locale is incompatible with operating system"),
errdetail("The database was initialized with LC_CTYPE \"%s\", "
" which is not recognized by setlocale().", ctype),
errhint("Recreate the database with another locale or install the missing locale.")));
/* Make the locale settings visible as GUC variables, too */
SetConfigOption("lc_collate", collate, PGC_INTERNAL, PGC_S_OVERRIDE);
SetConfigOption("lc_ctype", ctype, PGC_INTERNAL, PGC_S_OVERRIDE);
ReleaseSysCache(tup);
}
/*
* CheckMyDatabase -- fetch information from the pg_database entry for our DB
*/
static void
CheckMyDatabase(const char *name, bool am_superuser)
{
HeapTuple tup;
Form_pg_database dbform;
/* Fetch our pg_database row normally, via syscache */
tup = SearchSysCache(DATABASEOID,
ObjectIdGetDatum(MyDatabaseId),
0, 0, 0);
if (!HeapTupleIsValid(tup))
elog(ERROR, "cache lookup failed for database %u", MyDatabaseId);
dbform = (Form_pg_database) GETSTRUCT(tup);
/* This recheck is strictly paranoia */
if (strcmp(name, NameStr(dbform->datname)) != 0)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" has disappeared from pg_database",
name),
errdetail("Database OID %u now seems to belong to \"%s\".",
MyDatabaseId, NameStr(dbform->datname))));
/*
* Check permissions to connect to the database.
*
* These checks are not enforced when in standalone mode, so that there is
* a way to recover from disabling all access to all databases, for
* example "UPDATE pg_database SET datallowconn = false;".
*
* We do not enforce them for the autovacuum worker processes either.
*/
if (IsUnderPostmaster && !IsAutoVacuumWorkerProcess())
{
/*
* Check that the database is currently allowing connections.
*/
if (!dbform->datallowconn)
ereport(FATAL,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("database \"%s\" is not currently accepting connections",
name)));
/*
* Check privilege to connect to the database. (The am_superuser test
* is redundant, but since we have the flag, might as well check it
* and save a few cycles.)
*/
if (!am_superuser &&
pg_database_aclcheck(MyDatabaseId, GetUserId(),
ACL_CONNECT) != ACLCHECK_OK)
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("permission denied for database \"%s\"", name),
errdetail("User does not have CONNECT privilege.")));
/*
* Check connection limit for this database.
*
* There is a race condition here --- we create our PGPROC before
* checking for other PGPROCs. If two backends did this at about the
* same time, they might both think they were over the limit, while
* ideally one should succeed and one fail. Getting that to work
* exactly seems more trouble than it is worth, however; instead we
* just document that the connection limit is approximate.
*/
if (dbform->datconnlimit >= 0 &&
!am_superuser &&
CountDBBackends(MyDatabaseId) > dbform->datconnlimit)
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("too many connections for database \"%s\"",
name)));
}
/*
* OK, we're golden. Next to-do item is to save the encoding info out of
* the pg_database tuple.
*/
SetDatabaseEncoding(dbform->encoding);
/* Record it as a GUC internal option, too */
SetConfigOption("server_encoding", GetDatabaseEncodingName(),
PGC_INTERNAL, PGC_S_OVERRIDE);
/* If we have no other source of client_encoding, use server encoding */
SetConfigOption("client_encoding", GetDatabaseEncodingName(),
PGC_BACKEND, PGC_S_DEFAULT);
/* Use the right encoding in translated messages */
#ifdef ENABLE_NLS
pg_bind_textdomain_codeset(textdomain(NULL));
#endif
/*
* Lastly, set up any database-specific configuration variables.
*/
if (IsUnderPostmaster)
{
Datum datum;
bool isnull;
datum = SysCacheGetAttr(DATABASEOID, tup, Anum_pg_database_datconfig,
&isnull);
if (!isnull)
{
ArrayType *a = DatumGetArrayTypeP(datum);
/*
* We process all the options at SUSET level. We assume that the
* right to insert an option into pg_database was checked when it
* was inserted.
*/
ProcessGUCArray(a, PGC_SUSET, PGC_S_DATABASE, GUC_ACTION_SET);
}
}
ReleaseSysCache(tup);
}
/*
* lazy_vacuum_rel() -- perform LAZY VACUUM for one heap relation
*
* This routine vacuums a single heap, cleans out its indexes, and
* updates its relpages and reltuples statistics.
*
* At entry, we have already established a transaction and opened
* and locked the relation.
*/
void
lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
BufferAccessStrategy bstrategy)
{
LVRelStats *vacrelstats;
Relation *Irel;
int nindexes;
BlockNumber possibly_freeable;
PGRUsage ru0;
TimestampTz starttime = 0;
bool scan_all;
TransactionId freezeTableLimit;
pg_rusage_init(&ru0);
/* measure elapsed time iff autovacuum logging requires it */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration > 0)
starttime = GetCurrentTimestamp();
if (vacstmt->options & VACOPT_VERBOSE)
elevel = INFO;
else
elevel = DEBUG2;
vac_strategy = bstrategy;
vacuum_set_xid_limits(vacstmt->freeze_min_age, vacstmt->freeze_table_age,
onerel->rd_rel->relisshared,
&OldestXmin, &FreezeLimit, &freezeTableLimit);
scan_all = TransactionIdPrecedesOrEquals(onerel->rd_rel->relfrozenxid,
freezeTableLimit);
vacrelstats = (LVRelStats *) palloc0(sizeof(LVRelStats));
vacrelstats->old_rel_tuples = onerel->rd_rel->reltuples;
vacrelstats->num_index_scans = 0;
/* Open all indexes of the relation */
vac_open_indexes(onerel, RowExclusiveLock, &nindexes, &Irel);
vacrelstats->hasindex = (nindexes > 0);
/* Do the vacuuming */
lazy_scan_heap(onerel, vacrelstats, Irel, nindexes, scan_all);
/* Done with indexes */
vac_close_indexes(nindexes, Irel, NoLock);
/*
* Optionally truncate the relation.
*
* Don't even think about it unless we have a shot at releasing a goodly
* number of pages. Otherwise, the time taken isn't worth it.
*/
possibly_freeable = vacrelstats->rel_pages - vacrelstats->nonempty_pages;
if (possibly_freeable > 0 &&
(possibly_freeable >= REL_TRUNCATE_MINIMUM ||
possibly_freeable >= vacrelstats->rel_pages / REL_TRUNCATE_FRACTION))
lazy_truncate_heap(onerel, vacrelstats);
/* Vacuum the Free Space Map */
FreeSpaceMapVacuum(onerel);
/*
* Update statistics in pg_class. But don't change relfrozenxid if we
* skipped any pages.
*/
vac_update_relstats(onerel,
vacrelstats->rel_pages, vacrelstats->new_rel_tuples,
vacrelstats->hasindex,
(vacrelstats->scanned_pages < vacrelstats->rel_pages) ?
InvalidTransactionId :
FreezeLimit);
/* report results to the stats collector, too */
pgstat_report_vacuum(RelationGetRelid(onerel),
onerel->rd_rel->relisshared,
vacrelstats->new_rel_tuples);
/* and log the action if appropriate */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0)
{
if (Log_autovacuum_min_duration == 0 ||
TimestampDifferenceExceeds(starttime, GetCurrentTimestamp(),
Log_autovacuum_min_duration))
ereport(LOG,
(errmsg("automatic vacuum of table \"%s.%s.%s\": index scans: %d\n"
"pages: %d removed, %d remain\n"
"tuples: %.0f removed, %.0f remain\n"
"system usage: %s",
get_database_name(MyDatabaseId),
get_namespace_name(RelationGetNamespace(onerel)),
RelationGetRelationName(onerel),
vacrelstats->num_index_scans,
vacrelstats->pages_removed,
vacrelstats->rel_pages,
vacrelstats->tuples_deleted,
vacrelstats->new_rel_tuples,
pg_rusage_show(&ru0))));
}
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* The database can be specified by name, using the in_dbname parameter, or by
* OID, using the dboid parameter. In the latter case, the actual database
* name can be returned to the caller in out_dbname. If out_dbname isn't
* NULL, it must point to a buffer of size NAMEDATALEN.
*
* In bootstrap mode no parameters are used.
*
* The return value indicates whether the userID is a superuser. (That
* can only be tested inside a transaction, so we want to do it during
* the startup transaction rather than doing a separate one in postgres.c.)
*
* As of PostgreSQL 8.2, we expect InitProcess() was already called, so we
* already have a PGPROC struct ... but it's not filled in yet.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
void
InitPostgres(const char *in_dbname, Oid dboid, const char *username,
char *out_dbname)
{
bool bootstrap = IsBootstrapProcessingMode();
bool autovacuum = IsAutoVacuumWorkerProcess();
bool am_superuser;
char *fullpath;
char dbname[NAMEDATALEN];
/*
* Add my PGPROC struct to the ProcArray.
*
* Once I have done this, I am visible to other backends!
*/
InitProcessPhase2();
/* Initialize SessionState entry */
SessionState_Init();
/* Initialize memory protection */
GPMemoryProtect_Init();
#ifdef USE_ORCA
/* Initialize GPOPT */
InitGPOPT();
#endif
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data.
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
SharedInvalBackendInit(false);
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend id: %d", MyBackendId);
/* Now that we have a BackendId, we can participate in ProcSignal */
ProcSignalInit(MyBackendId);
/*
* bufmgr needs another initialization call too
*/
InitBufferPoolBackend();
/*
* Initialize local process's access to XLOG. In bootstrap case we may
* skip this since StartupXLOG() was run instead.
*/
if (!bootstrap)
InitXLOGAccess();
/*
* Initialize the relation cache and the system catalog caches. Note that
* no catalog access happens here; we only set up the hashtable structure.
* We must do this before starting a transaction because transaction abort
* would try to touch these hashtables.
*/
RelationCacheInitialize();
InitCatalogCache();
InitPlanCache();
/* Initialize portal manager */
EnablePortalManager();
/* Initialize stats collection --- must happen before first xact */
if (!bootstrap)
pgstat_initialize();
/*
* Load relcache entries for the shared system catalogs. This must create
* at least entries for pg_database and catalogs used for authentication.
*/
RelationCacheInitializePhase2();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This has to
* be after we've initialized all the low-level modules like the buffer
* manager, because during shutdown this has to run before the low-level
* modules start to close down. On the other hand, we want it in place
* before we begin our first transaction --- if we fail during the
* initialization transaction, as is entirely possible, we need the
* AbortTransaction call to clean up.
*/
on_shmem_exit(ShutdownPostgres, 0);
/* TODO: autovacuum launcher should be done here? */
/*
* Start a new transaction here before first access to db, and get a
* snapshot. We don't have a use for the snapshot itself, but we're
* interested in the secondary effect that it sets RecentGlobalXmin.
*/
if (!bootstrap)
{
StartTransactionCommand();
(void) GetTransactionSnapshot();
}
/*
* Figure out our postgres user id, and see if we are a superuser.
*
* In standalone mode and in the autovacuum process, we use a fixed id,
* otherwise we figure it out from the authenticated user name.
*/
if (bootstrap || autovacuum)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
am_superuser = true;
if (!ThereIsAtLeastOneRole())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no roles are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" CREATEUSER;.",
username)));
}
else
{
/* normal multiuser case */
Assert(MyProcPort != NULL);
PerformAuthentication(MyProcPort);
InitializeSessionUserId(username);
am_superuser = superuser();
}
/*
* Check a normal user hasn't connected to a superuser reserved slot.
*/
if (!am_superuser &&
ReservedBackends > 0 &&
!HaveNFreeProcs(ReservedBackends))
ereport(FATAL,
(errcode(ERRCODE_TOO_MANY_CONNECTIONS),
errmsg("connection limit exceeded for non-superusers"),
errSendAlert(true)));
/*
* If walsender, we don't want to connect to any particular database. Just
* finish the backend startup by processing any options from the startup
* packet, and we're done.
*/
if (am_walsender)
{
Assert(!bootstrap);
/*
* We don't have replication role, which existed in postgres.
*/
if (!superuser())
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser role to start walsender")));
/* process any options passed in the startup packet */
if (MyProcPort != NULL)
process_startup_options(MyProcPort, am_superuser);
/* Apply PostAuthDelay as soon as we've read all options */
if (PostAuthDelay > 0)
pg_usleep(PostAuthDelay * 1000000L);
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
pgstat_bestart();
/* close the transaction we started above */
CommitTransactionCommand();
return;
}
/*
* Set up the global variables holding database id and path. But note we
* won't actually try to touch the database just yet.
*
* We take a shortcut in the bootstrap case, otherwise we have to look up
* the db name in pg_database.
*/
if (bootstrap)
{
MyDatabaseId = TemplateDbOid;
MyDatabaseTableSpace = DEFAULTTABLESPACE_OID;
}
else if (in_dbname != NULL)
{
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTuple(in_dbname);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", in_dbname)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
/* take database name from the caller, just for paranoia */
strlcpy(dbname, in_dbname, sizeof(dbname));
pfree(tuple);
}
else
{
/* caller specified database by OID */
HeapTuple tuple;
Form_pg_database dbform;
tuple = GetDatabaseTupleByOid(dboid);
if (!HeapTupleIsValid(tuple))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database %u does not exist", dboid)));
dbform = (Form_pg_database) GETSTRUCT(tuple);
MyDatabaseId = HeapTupleGetOid(tuple);
MyDatabaseTableSpace = dbform->dattablespace;
Assert(MyDatabaseId == dboid);
strlcpy(dbname, NameStr(dbform->datname), sizeof(dbname));
/* pass the database name back to the caller */
if (out_dbname)
strcpy(out_dbname, dbname);
pfree(tuple);
}
/* Now we can mark our PGPROC entry with the database ID */
/* (We assume this is an atomic store so no lock is needed) */
MyProc->databaseId = MyDatabaseId;
/*
* Now, take a writer's lock on the database we are trying to connect to.
* If there is a concurrently running DROP DATABASE on that database, this
* will block us until it finishes (and has committed its update of
* pg_database).
*
* Note that the lock is not held long, only until the end of this startup
* transaction. This is OK since we are already advertising our use of
* the database in the PGPROC array; anyone trying a DROP DATABASE after
* this point will see us there.
*
* Note: use of RowExclusiveLock here is reasonable because we envision
* our session as being a concurrent writer of the database. If we had a
* way of declaring a session as being guaranteed-read-only, we could use
* AccessShareLock for such sessions and thereby not conflict against
* CREATE DATABASE.
*/
if (!bootstrap)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
RowExclusiveLock);
/*
* Recheck pg_database to make sure the target database hasn't gone away.
* If there was a concurrent DROP DATABASE, this ensures we will die
* cleanly without creating a mess.
*/
if (!bootstrap)
{
HeapTuple tuple;
tuple = GetDatabaseTuple(dbname);
if (!HeapTupleIsValid(tuple) ||
MyDatabaseId != HeapTupleGetOid(tuple) ||
MyDatabaseTableSpace != ((Form_pg_database) GETSTRUCT(tuple))->dattablespace)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist", dbname),
errdetail("It seems to have just been dropped or renamed.")));
}
fullpath = GetDatabasePath(MyDatabaseId, MyDatabaseTableSpace);
if (!bootstrap)
{
if (access(fullpath, F_OK) == -1)
{
if (errno == ENOENT)
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname),
errdetail("The database subdirectory \"%s\" is missing.",
fullpath)));
else
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not access directory \"%s\": %m",
fullpath)));
}
ValidatePgVersion(fullpath);
}
SetDatabasePath(fullpath);
/*
* It's now possible to do real access to the system catalogs.
*
* Load relcache entries for the system catalogs. This must create at
* least the minimum set of "nailed-in" cache entries.
*/
RelationCacheInitializePhase3();
/*
* Now we have full access to catalog including toast tables,
* we can process pg_authid.rolconfig. This ought to come before
* processing startup options so that it can override the settings.
*/
if (!bootstrap)
ProcessRoleGUC();
/* set up ACL framework (so CheckMyDatabase can check permissions) */
initialize_acl();
/*
* Re-read the pg_database row for our database, check permissions and set
* up database-specific GUC settings. We can't do this until all the
* database-access infrastructure is up. (Also, it wants to know if the
* user is a superuser, so the above stuff has to happen first.)
*/
if (!bootstrap)
CheckMyDatabase(dbname, am_superuser);
/*
* Now process any command-line switches and any additional GUC variable
* settings passed in the startup packet. We couldn't do this before
* because we didn't know if client is a superuser.
*/
if (MyProcPort != NULL)
process_startup_options(MyProcPort, am_superuser);
/*
* Maintenance Mode: allow superuser to connect when
* gp_maintenance_conn GUC is set. We cannot check it until
* process_startup_options parses the GUC.
*/
if (gp_maintenance_mode && Gp_role == GP_ROLE_DISPATCH &&
!(superuser() && gp_maintenance_conn))
ereport(FATAL,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("maintenance mode: connected by superuser only"),
errSendAlert(false)));
/*
* MPP: If we were started in utility mode then we only want to allow
* incoming sessions that specify gp_session_role=utility as well. This
* lets the bash scripts start the QD in utility mode and connect in but
* protect ourselves from normal clients who might be trying to connect to
* the system while we startup.
*/
if ((Gp_role == GP_ROLE_UTILITY) && (Gp_session_role != GP_ROLE_UTILITY))
{
ereport(FATAL,
(errcode(ERRCODE_CANNOT_CONNECT_NOW),
errmsg("System was started in master-only utility mode - only utility mode connections are allowed")));
}
/* Apply PostAuthDelay as soon as we've read all options */
if (PostAuthDelay > 0)
pg_usleep(PostAuthDelay * 1000000L);
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/*
* MPP package setup
*
* Primary function is to establish connctions to the qExecs.
* This is SKIPPED when the database is in bootstrap mode or
* Is not UnderPostmaster.
*/
if (!bootstrap && IsUnderPostmaster)
{
cdb_setup();
on_proc_exit( cdb_cleanup, 0 );
}
/*
* MPP SharedSnapshot Setup
*/
if (Gp_role == GP_ROLE_DISPATCH)
{
addSharedSnapshot("Query Dispatcher", gp_session_id);
}
else if (Gp_role == GP_ROLE_DISPATCHAGENT)
{
SharedLocalSnapshotSlot = NULL;
}
else if (Gp_segment == -1 && Gp_role == GP_ROLE_EXECUTE && !Gp_is_writer)
{
/*
* Entry db singleton QE is a user of the shared snapshot -- not a creator.
* The lookup will occur once the distributed snapshot has been received.
*/
lookupSharedSnapshot("Entry DB Singleton", "Query Dispatcher", gp_session_id);
}
else if (Gp_role == GP_ROLE_EXECUTE)
{
if (Gp_is_writer)
{
addSharedSnapshot("Writer qExec", gp_session_id);
}
else
{
/*
* NOTE: This assumes that the Slot has already been
* allocated by the writer. Need to make sure we
* always allocate the writer qExec first.
*/
lookupSharedSnapshot("Reader qExec", "Writer qExec", gp_session_id);
}
}
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
return;
}
