/* --------------------------------
* 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();
/* 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;
}
static void
QDMirroringUpdate(
QDMIRRORUpdateMask updateMask,
bool validFlag,
QDMIRRORState state,
QDMIRRORDisabledReason disabledReason,
struct timeval *lastLogTimeVal,
char *errorMessage)
{
#define UPDATE_VALIDFLAG_CMD "update gp_configuration set valid='%c' where dbid = CAST(%d AS SMALLINT)"
#define UPDATE_MASTER_MIRRORING_CMD "update gp_master_mirroring set (summary_state, detail_state, log_time, error_message) = ('%s', %s, '%s'::timestamptz, %s);"
int count = 0;
char cmd[200 + QDMIRRORErrorMessageSize * 2 + 3];
char detailValue[100];
char logTimeStr[128];
char *summaryStateString;
char *detailStateString;
char errorMessageQuoted[QDMIRRORErrorMessageSize * 2 + 3];
char *user;
MemoryContext mcxt = CurrentMemoryContext;
Segment *master = NULL;
volatile PQExpBuffer entryBuffer = NULL;
volatile PGconn *entryConn = NULL;
volatile PGresult *rs = NULL;
PG_TRY();
{
StartTransactionCommand();
user = getDBSuperuserName("QDMirroringUpdate");
Assert(user != NULL);
master = GetMasterSegment();
entryBuffer = createPQExpBuffer();
if (PQExpBufferBroken(entryBuffer))
{
destroyPQExpBuffer(entryBuffer);
ereport(ERROR, (errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("QDMirroringUpdate: out of memory")));
/* not reached. */
}
/*
* initialize libpq connection buffer, we only need to initialize it
* once.
*/
initPQConnectionBuffer(master, user, NULL, entryBuffer, true);
FreeSegment(master);
free(user);
/*
* Call libpq to connect
*/
entryConn = PQconnectdb(entryBuffer->data);
if (PQstatus((PGconn *)entryConn) == CONNECTION_BAD)
{
/*
* When we get an error, we strdup it here. When the main thread
* checks for errors, it makes a palloc copy of this, and frees
* this.
*/
char *error_message = strdup(PQerrorMessage((PGconn *)entryConn));
if (!error_message)
{
ereport(ERROR, (errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("QDMirroringUpdate: out of memory")));
}
destroyPQExpBuffer(entryBuffer);
PQfinish((PGconn *)entryConn);
entryConn = NULL;
elog(FATAL, "QDMirroringUpdate: setting segDB state failed, error connecting to entry db, error: %s", error_message);
}
/* finally, we're ready to actually get some stuff done. */
do
{
rs = PQexec((PGconn *)entryConn, "BEGIN");
if (PQresultStatus((PGresult *)rs) != PGRES_COMMAND_OK)
break;
if ((updateMask & QDMIRROR_UPDATEMASK_VALIDFLAG) != 0)
{
count = snprintf(cmd, sizeof(cmd), UPDATE_VALIDFLAG_CMD, (validFlag ? 't' : 'f'), ftsQDMirrorInfo->dbid);
if (count >= sizeof(cmd))
{
ereport(ERROR, (errcode(ERRCODE_INTERNAL_ERROR),
errmsg("QDMirroringUpdate: format command string failure")));
}
rs = PQexec((PGconn *)entryConn, cmd);
if (PQresultStatus((PGresult *)rs) != PGRES_COMMAND_OK)
{
ereport(ERROR, (errcode(ERRCODE_INTERNAL_ERROR),
errmsg("QDMirroringUpdate: could not execute command '%s'", cmd)));
break;
}
}
if ((updateMask & QDMIRROR_UPDATEMASK_MASTERMIRRORING)!= 0)
{
switch (state)
{
case QDMIRROR_STATE_NONE:
summaryStateString = "None";
break;
case QDMIRROR_STATE_NOTCONFIGURED:
summaryStateString = "Not Configured";
break;
case QDMIRROR_STATE_CONNECTINGWALSENDSERVER:
case QDMIRROR_STATE_POSITIONINGTOEND:
case QDMIRROR_STATE_CATCHUPPENDING:
case QDMIRROR_STATE_CATCHINGUP:
summaryStateString = "Synchronizing";
break;
case QDMIRROR_STATE_SYNCHRONIZED:
summaryStateString = "Synchronized";
break;
case QDMIRROR_STATE_DISABLED:
summaryStateString = "Not Synchronized";
break;
default:
summaryStateString = "Unknown";
break;
}
if (state == QDMIRROR_STATE_DISABLED)
{
detailStateString =
QDMirroringDisabledReasonToString(disabledReason);
}
else
{
detailStateString = NULL;
}
if (detailStateString == NULL)
{
strcpy(detailValue, "null");
}
else
{
count = snprintf(detailValue, sizeof(detailValue), "'%s'", detailStateString);
if (count >= sizeof(detailValue))
{
ereport(ERROR, (errcode(ERRCODE_INTERNAL_ERROR),
errmsg("QDMirroringUpdate: format command string failure")));
}
}
QDMirroringFormatTime(logTimeStr, sizeof(logTimeStr), lastLogTimeVal);
/*
* Escape quote the error string before putting in DML statement...
*/
if (errorMessage != NULL)
{
int errorMessageLen = strlen(errorMessage);
if (errorMessageLen == 0)
{
strcpy(errorMessageQuoted, "null");
}
else
{
size_t escapedLen;
errorMessageQuoted[0] = '\'';
escapedLen = PQescapeString(&errorMessageQuoted[1], errorMessage, errorMessageLen);
errorMessageQuoted[escapedLen + 1] = '\'';
errorMessageQuoted[escapedLen + 2] = '\0';
elog((Debug_print_qd_mirroring ? LOG : DEBUG5), "Error message quoted: \"%s\"", errorMessageQuoted);
}
}
else
{
strcpy(errorMessageQuoted, "null");
}
count = snprintf(cmd, sizeof(cmd), UPDATE_MASTER_MIRRORING_CMD,
summaryStateString, detailValue, logTimeStr, errorMessageQuoted);
if (count >= sizeof(cmd))
{
ereport(ERROR, (errcode(ERRCODE_INTERNAL_ERROR),
errmsg("QDMirroringUpdate: format command string failure")));
}
rs = PQexec((PGconn *)entryConn, cmd);
if (PQresultStatus((PGresult *)rs) != PGRES_COMMAND_OK)
{
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("QDMirroringUpdate: could not execute command '%s'", cmd)));
break;
}
elog((Debug_print_qd_mirroring ? LOG : DEBUG5),
"Successfully executed command \"%s\"", cmd);
rs = PQexec((PGconn *)entryConn, "COMMIT");
if (PQresultStatus((PGresult *)rs) != PGRES_COMMAND_OK)
break;
}
}
while (0);
PQclear((PGresult *)rs);
PQfinish((PGconn *)entryConn);
destroyPQExpBuffer(entryBuffer);
CommitTransactionCommand();
}
PG_CATCH();
{
PQclear((PGresult *)rs);
PQfinish((PGconn *)entryConn);
destroyPQExpBuffer(entryBuffer);
AbortCurrentTransaction();
}
PG_END_TRY();
MemoryContextSwitchTo(mcxt); /* Just incase we hit an error */
return;
}
/*
* 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)));
/*
* 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_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;
/*
* 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();
}
}
}
}
/* --------------------------------
* 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.
*
* Similarly, the username can be passed by name, using the username parameter,
* or by OID using the useroid parameter.
*
* 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,
Oid useroid, char *out_dbname, bool override_allow_connections)
{
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, CheckDeadLockAlert);
RegisterTimeout(STATEMENT_TIMEOUT, StatementTimeoutHandler);
RegisterTimeout(LOCK_TIMEOUT, LockTimeoutHandler);
RegisterTimeout(IDLE_IN_TRANSACTION_SESSION_TIMEOUT,
IdleInTransactionSessionTimeoutHandler);
}
/*
* 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.
*
* We don't yet have an aux-process resource owner, but StartupXLOG
* and ShutdownXLOG will need one. Hence, create said resource owner
* (and register a callback to clean it up after ShutdownXLOG runs).
*/
CreateAuxProcessResourceOwner();
StartupXLOG();
/* Release (and warn about) any buffer pins leaked in StartupXLOG */
ReleaseAuxProcessResources(true);
/* Reset CurrentResourceOwner to nothing for the moment */
CurrentResourceOwner = NULL;
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())
{
/* report this backend in the PgBackendStatus array */
pgstat_bestart();
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 != NULL ? username : "******")));
}
else if (IsBackgroundWorker)
{
if (username == NULL && !OidIsValid(useroid))
{
InitializeSessionUserIdStandalone();
am_superuser = true;
}
else
{
InitializeSessionUserId(username, useroid);
am_superuser = superuser();
}
}
else
{
/* normal multiuser case */
Assert(MyProcPort != NULL);
PerformAuthentication(MyProcPort);
InitializeSessionUserId(username, useroid);
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 connection slots are reserved for superusers. Replication
* connections are drawn from slots reserved with max_wal_senders and not
* limited by max_connections or superuser_reserved_connections.
*/
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 = dbform->oid;
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 = dbform->oid;
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)
{
pgstat_bestart();
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 != ((Form_pg_database) GETSTRUCT(tuple))->oid ||
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, override_allow_connections);
/*
* 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();
/* Initialize this backend's session state. */
InitializeSession();
/* report this backend in the PgBackendStatus array */
if (!bootstrap)
pgstat_bestart();
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
}
static config_log_objects *
initialize_objects(void)
{
config_log_objects *objects;
int ret;
int ntup;
bool isnull;
StringInfoData buf;
objects = palloc(sizeof(config_log_objects));
objects->table_name = pstrdup("pg_settings_log");
objects->function_name = pstrdup("pg_settings_logger");
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "Verifying config log objects");
initStringInfo(&buf);
appendStringInfo(
&buf,
"SELECT COUNT(*)\
FROM information_schema.tables\
WHERE table_schema='%s'\
AND table_name ='%s'\
AND table_type='BASE TABLE'",
config_log_schema,
objects->table_name
);
ret = SPI_execute(buf.data, true, 0);
if (ret != SPI_OK_SELECT)
{
ereport(FATAL,
(errmsg("SPI_execute failed: SPI error code %d", ret)
));
}
/* This should never happen */
if (SPI_processed != 1)
{
elog(FATAL, "not a singleton result");
}
ntup = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
/* This should never happen */
if (isnull)
{
elog(FATAL, "null result");
}
if (ntup == 0)
{
ereport(FATAL,
(
errmsg("Expected config log table '%s.%s' not found", config_log_schema,
objects->table_name),
errhint("Ensure superuser search_path includes the schema used by config_log; "
"check config_log.* GUC settings")
));
}
/* check function pg_settings_logger() exists */
resetStringInfo(&buf);
appendStringInfo(
&buf,
"SELECT COUNT(*) FROM pg_catalog.pg_proc p \
INNER JOIN pg_catalog.pg_namespace n ON n.oid = p.pronamespace \
WHERE p.proname='%s' \
AND n.nspname='%s' \
AND p.pronargs = 0",
objects->function_name,
config_log_schema
);
ret = SPI_execute(buf.data, true, 0);
if (ret != SPI_OK_SELECT)
{
ereport(FATAL,
(errmsg("SPI_execute failed: SPI error code %d", ret)));
}
if (SPI_processed != 1)
{
elog(FATAL, "not a singleton result");
}
ntup = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (isnull)
{
elog(FATAL, "null result");
}
if (ntup == 0)
{
ereport(FATAL,
(
errmsg("Expected config log function '%s.%s' not found", config_log_schema,
objects->function_name),
errhint("Ensure superuser search_path includes the schema used by config_log; "
"check config_log.* GUC settings")
));
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
log_info("initialized, database objects validated");
/* execute pg_settings_logger() here to catch any settings which have changed after server restart */
execute_pg_settings_logger(objects);
return objects;
}
static void
SaveBuffers(void)
{
int i;
int num_buffers;
int log_level = DEBUG3;
SavedBuffer *saved_buffers;
volatile BufferDesc *bufHdr; // XXX: Do we really need volatile here?
FILE *file = NULL;
int database_counter= 0;
Oid prev_database = InvalidOid;
Oid prev_filenode = InvalidOid;
ForkNumber prev_forknum = InvalidForkNumber;
BlockNumber prev_blocknum = InvalidBlockNumber;
BlockNumber range_counter = 0;
const char *savefile_path;
/*
* XXX: If the memory request fails, ask for a smaller memory chunk, and use
* it to create chunks of save-files, and make the workers read those chunks.
*
* This is not a concern as of now, so deferred; there's at least one other
* place that allocates (NBuffers * (much_bigger_struct)), so this seems to
* be an acceptable practice.
*/
saved_buffers = (SavedBuffer *) palloc(sizeof(SavedBuffer) * NBuffers);
/* Lock the buffer partitions for reading. */
for (i = 0; i < NUM_BUFFER_PARTITIONS; ++i)
LWLockAcquire(FirstBufMappingLock + i, LW_SHARED);
/* Scan and save a list of valid buffers. */
for (num_buffers = 0, i = 0, bufHdr = BufferDescriptors; i < NBuffers; ++i, ++bufHdr)
{
/* Lock each buffer header before inspecting. */
LockBufHdr(bufHdr);
/* Skip invalid buffers */
if ((bufHdr->flags & BM_VALID) && (bufHdr->flags & BM_TAG_VALID))
{
saved_buffers[num_buffers].database = bufHdr->tag.rnode.dbNode;
saved_buffers[num_buffers].filenode = bufHdr->tag.rnode.relNode;
saved_buffers[num_buffers].forknum = bufHdr->tag.forkNum;
saved_buffers[num_buffers].blocknum = bufHdr->tag.blockNum;
++num_buffers;
}
UnlockBufHdr(bufHdr);
}
/* Unlock the buffer partitions in reverse order, to avoid a deadlock. */
for (i = NUM_BUFFER_PARTITIONS - 1; i >= 0; --i)
LWLockRelease(FirstBufMappingLock + i);
/*
* Sort the list, so that we can optimize the storage of these buffers.
*
* The side-effect of this storage optimization is that when reading the
* blocks back from relation forks, it leads to sequential reads, which
* improve the restore speeds quite considerably as compared to random reads
* from different blocks all over the data directory.
*/
pg_qsort(saved_buffers, num_buffers, sizeof(SavedBuffer), SavedBufferCmp);
/* Connect to the database and start a transaction for database name lookups. */
BackgroundWorkerInitializeConnection(guc_default_database, NULL);
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "saving buffers");
for (i = 0; i < num_buffers; ++i)
{
int j;
SavedBuffer *buf = &saved_buffers[i];
if (i == 0)
{
/*
* Special case for global objects. The sort brings them to the
* front of the list.
*/
/* Make sure the first buffer we save belongs to global object. */
Assert(buf->database == InvalidOid);
/*
* Database number (and save-file name) 1 is reserverd for storing
* list of buffers of global objects.
*/
database_counter = 1;
savefile_path = getSavefileName(database_counter);
file = fileOpen(savefile_path, PG_BINARY_W);
writeDBName("", file, savefile_path);
prev_database = buf->database;
}
if (buf->database != prev_database)
{
char *dbname;
/*
* We are beginning to process a different database than the
* previous one; close the save-file of previous database, and open
* a new one.
*/
++database_counter;
dbname = get_database_name(buf->database);
Assert(dbname != NULL);
if (file != NULL)
fileClose(file, savefile_path);
savefile_path = getSavefileName(database_counter);
file = fileOpen(savefile_path, PG_BINARY_W);
writeDBName(dbname, file, savefile_path);
pfree(dbname);
/* Reset trackers appropriately */
prev_database = buf->database;
prev_filenode = InvalidOid;
prev_forknum = InvalidForkNumber;
prev_blocknum = InvalidBlockNumber;
range_counter = 0;
}
if (buf->filenode != prev_filenode)
{
/* We're beginning to process a new relation; emit a record for it. */
fileWrite("r", 1, file, savefile_path);
fileWrite(&(buf->filenode), sizeof(Oid), file, savefile_path);
/* Reset trackers appropriately */
prev_filenode = buf->filenode;
prev_forknum = InvalidForkNumber;
prev_blocknum = InvalidBlockNumber;
range_counter = 0;
}
if (buf->forknum != prev_forknum)
{
/*
* We're beginning to process a new fork of this relation; add a
* record for it.
*/
fileWrite("f", 1, file, savefile_path);
fileWrite(&(buf->forknum), sizeof(ForkNumber), file, savefile_path);
/* Reset trackers appropriately */
prev_forknum = buf->forknum;
prev_blocknum = InvalidBlockNumber;
range_counter = 0;
}
ereport(log_level,
(errmsg("writer: writing block db %d filenode %d forknum %d blocknum %d",
database_counter, prev_filenode, prev_forknum, buf->blocknum)));
fileWrite("b", 1, file, savefile_path);
fileWrite(&(buf->blocknum), sizeof(BlockNumber), file, savefile_path);
prev_blocknum = buf->blocknum;
/*
* If a continuous range of blocks follows this block, then emit one
* entry for the range, instead of one for each block.
*/
range_counter = 0;
for ( j = i+1; j < num_buffers; ++j)
{
SavedBuffer *tmp = &saved_buffers[j];
if (tmp->database == prev_database
&& tmp->filenode == prev_filenode
&& tmp->forknum == prev_forknum
&& tmp->blocknum == (prev_blocknum + range_counter + 1))
{
++range_counter;
}
}
if (range_counter != 0)
{
ereport(log_level,
(errmsg("writer: writing range db %d filenode %d forknum %d blocknum %d range %d",
database_counter, prev_filenode, prev_forknum, prev_blocknum, range_counter)));
fileWrite("N", 1, file, savefile_path);
fileWrite(&range_counter, sizeof(range_counter), file, savefile_path);
i += range_counter;
}
}
ereport(LOG,
(errmsg("Buffer Saver: saved metadata of %d blocks", num_buffers)));
Assert(file != NULL);
fileClose(file, savefile_path);
pfree(saved_buffers);
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
}
void
worker_spi_main(Datum main_arg)
{
int index = DatumGetInt32(main_arg);
worktable *table;
StringInfoData buf;
char name[20];
table = palloc(sizeof(worktable));
sprintf(name, "schema%d", index);
table->schema = pstrdup(name);
table->name = pstrdup("counted");
/* Establish signal handlers before unblocking signals. */
pqsignal(SIGHUP, worker_spi_sighup);
pqsignal(SIGTERM, worker_spi_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection("postgres", NULL);
elog(LOG, "%s initialized with %s.%s",
MyBgworkerEntry->bgw_name, table->schema, table->name);
initialize_worker_spi(table);
/*
* Quote identifiers passed to us. Note that this must be done after
* initialize_worker_spi, because that routine assumes the names are not
* quoted.
*
* Note some memory might be leaked here.
*/
table->schema = quote_identifier(table->schema);
table->name = quote_identifier(table->name);
initStringInfo(&buf);
appendStringInfo(&buf,
"WITH deleted AS (DELETE "
"FROM %s.%s "
"WHERE type = 'delta' RETURNING value), "
"total AS (SELECT coalesce(sum(value), 0) as sum "
"FROM deleted) "
"UPDATE %s.%s "
"SET value = %s.value + total.sum "
"FROM total WHERE type = 'total' "
"RETURNING %s.value",
table->schema, table->name,
table->schema, table->name,
table->name,
table->name);
/*
* Main loop: do this until the SIGTERM handler tells us to terminate
*/
while (!got_sigterm)
{
int ret;
int rc;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(MyLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
worker_spi_naptime * 1000L,
PG_WAIT_EXTENSION);
ResetLatch(MyLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
/*
* In case of a SIGHUP, just reload the configuration.
*/
if (got_sighup)
{
got_sighup = false;
ProcessConfigFile(PGC_SIGHUP);
}
/*
* Start a transaction on which we can run queries. Note that each
* StartTransactionCommand() call should be preceded by a
* SetCurrentStatementStartTimestamp() call, which sets both the time
* for the statement we're about the run, and also the transaction
* start time. Also, each other query sent to SPI should probably be
* preceded by SetCurrentStatementStartTimestamp(), so that statement
* start time is always up to date.
*
* The SPI_connect() call lets us run queries through the SPI manager,
* and the PushActiveSnapshot() call creates an "active" snapshot
* which is necessary for queries to have MVCC data to work on.
*
* The pgstat_report_activity() call makes our activity visible
* through the pgstat views.
*/
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, buf.data);
/* We can now execute queries via SPI */
ret = SPI_execute(buf.data, false, 0);
if (ret != SPI_OK_UPDATE_RETURNING)
elog(FATAL, "cannot select from table %s.%s: error code %d",
table->schema, table->name, ret);
if (SPI_processed > 0)
{
bool isnull;
int32 val;
val = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
if (!isnull)
elog(LOG, "%s: count in %s.%s is now %d",
MyBgworkerEntry->bgw_name,
table->schema, table->name, val);
}
/*
* And finish our transaction.
*/
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_stat(false);
pgstat_report_activity(STATE_IDLE, NULL);
}
proc_exit(1);
}
/*
* ProcessCatchupEvent
*
* Respond to a catchup event (PROCSIG_CATCHUP_INTERRUPT) from another
* backend.
*
* This is called either directly from the PROCSIG_CATCHUP_INTERRUPT
* signal handler, or the next time control reaches the outer idle loop
* (assuming there's still anything to do by then).
*/
static void
ProcessCatchupEvent(void)
{
bool notify_enabled;
bool client_wait_timeout_enabled;
DtxContext saveDistributedTransactionContext;
/*
* Funny indentation to keep the code inside identical to upstream
* while at the same time supporting CMockery which has problems with
* multiple bracing on column 1.
*/
PG_TRY();
{
in_process_catchup_event = 1;
/* Must prevent SIGUSR2 and SIGALRM(for IdleSessionGangTimeout) interrupt while I am running */
notify_enabled = DisableNotifyInterrupt();
client_wait_timeout_enabled = DisableClientWaitTimeoutInterrupt();
/*
* What we need to do here is cause ReceiveSharedInvalidMessages() to run,
* which will do the necessary work and also reset the
* catchupInterruptOccurred flag. If we are inside a transaction we can
* just call AcceptInvalidationMessages() to do this. If we aren't, we
* start and immediately end a transaction; the call to
* AcceptInvalidationMessages() happens down inside transaction start.
*
* It is awfully tempting to just call AcceptInvalidationMessages()
* without the rest of the xact start/stop overhead, and I think that
* would actually work in the normal case; but I am not sure that things
* would clean up nicely if we got an error partway through.
*/
if (IsTransactionOrTransactionBlock())
{
elog(DEBUG1, "ProcessCatchupEvent inside transaction");
AcceptInvalidationMessages();
}
else
{
elog(DEBUG1, "ProcessCatchupEvent outside transaction");
/*
* Save distributed transaction context first.
*/
saveDistributedTransactionContext = DistributedTransactionContext;
DistributedTransactionContext = DTX_CONTEXT_LOCAL_ONLY;
StartTransactionCommand();
CommitTransactionCommand();
DistributedTransactionContext = saveDistributedTransactionContext;
}
if (notify_enabled)
EnableNotifyInterrupt();
if (client_wait_timeout_enabled)
EnableClientWaitTimeoutInterrupt();
in_process_catchup_event = 0;
}
PG_CATCH();
{
in_process_catchup_event = 0;
PG_RE_THROW();
}
PG_END_TRY();
}
static void
ReadBlocks(int filenum)
{
FILE *file;
char record_type;
char *dbname;
Oid record_filenode;
ForkNumber record_forknum;
BlockNumber record_blocknum;
BlockNumber record_range;
int log_level = DEBUG3;
Oid relOid = InvalidOid;
Relation rel = NULL;
bool skip_relation = false;
bool skip_fork = false;
bool skip_block = false;
BlockNumber nblocks = 0;
BlockNumber blocks_restored = 0;
const char *filepath;
/*
* If this condition changes, then this code, and the code in the writer
* will need to be changed; especially the format specifiers in log and
* error messages.
*/
StaticAssertStmt(MaxBlockNumber == 0xFFFFFFFE, "Code may need review.");
filepath = getSavefileName(filenum);
file = fileOpen(filepath, PG_BINARY_R);
dbname = readDBName(file, filepath);
/*
* When restoring global objects, the dbname is zero-length string, and non-
* zero length otherwise. And filenum is never expected to be smaller than 1.
*/
Assert(filenum >= 1);
Assert(filenum == 1 ? strlen(dbname) == 0 : strlen(dbname) > 0);
/* To restore the global objects, use default database */
BackgroundWorkerInitializeConnection(filenum == 1 ? guc_default_database : dbname, NULL);
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "restoring buffers");
/*
* Note that in case of a read error, we will leak relcache entry that we may
* currently have open. In case of EOF, we close the relation after the loop.
*/
while (fileRead(&record_type, 1, file, true, filepath))
{
/*
* If we want to process the signals, this seems to be the best place
* to do it. Generally the backends refrain from processing config file
* while in transaction, but that's more for the fear of allowing GUC
* changes to affect expression evaluation, causing different results
* for the same expression in a transaction. Since this worker is not
* processing any queries, it is okay to process the config file here.
*
* Even though it's okay to process SIGHUP here, doing so doesn't add
* any value. The only reason we might want to process config file here
* would be to allow the user to interrupt the BlockReader's operation
* by changing this extenstion's GUC parameter. But the user can do that
* anyway, using SIGTERM or pg_terminate_backend().
*/
/* Stop processing the save-file if the Postmaster wants us to die. */
if (got_sigterm)
break;
ereport(log_level,
(errmsg("record type %x - %c", record_type, record_type)));
switch (record_type)
{
case 'r':
{
/* Close the previous relation, if any. */
if (rel)
{
relation_close(rel, AccessShareLock);
rel = NULL;
}
record_forknum = InvalidForkNumber;
record_blocknum = InvalidBlockNumber;
nblocks = 0;
fileRead(&record_filenode, sizeof(Oid), file, false, filepath);
relOid = GetRelOid(record_filenode);
ereport(log_level, (errmsg("processing filenode %u, relation %u",
record_filenode, relOid)));
/*
* If the relation has been rewritten/dropped since we saved it,
* just skip it and process the next relation.
*/
if (relOid == InvalidOid)
skip_relation = true;
else
{
skip_relation = false;
/* Open the relation */
rel = relation_open(relOid, AccessShareLock);
RelationOpenSmgr(rel);
}
}
break;
case 'f':
{
record_blocknum = InvalidBlockNumber;
nblocks = 0;
fileRead(&record_forknum, sizeof(ForkNumber), file, false, filepath);
if (skip_relation)
continue;
if (rel == NULL)
ereport(ERROR,
(errmsg("found a fork record without a preceeding relation record")));
ereport(log_level, (errmsg("processing fork %d", record_forknum)));
if (!smgrexists(rel->rd_smgr, record_forknum))
skip_fork = true;
else
{
skip_fork = false;
nblocks = RelationGetNumberOfBlocksInFork(rel, record_forknum);
}
}
break;
case 'b':
{
if (record_forknum == InvalidForkNumber)
ereport(ERROR,
(errmsg("found a block record without a preceeding fork record")));
fileRead(&record_blocknum, sizeof(BlockNumber), file, false, filepath);
if (skip_relation || skip_fork)
continue;
/*
* Don't try to read past the file; the file may have been shrunk
* by a vaccum/truncate operation.
*/
if (record_blocknum >= nblocks)
{
ereport(log_level,
(errmsg("reader %d skipping block filenode %u forknum %d blocknum %u",
filenum, record_filenode, record_forknum, record_blocknum)));
skip_block = true;
continue;
}
else
{
Buffer buf;
skip_block = false;
ereport(log_level,
(errmsg("reader %d reading block filenode %u forknum %d blocknum %u",
filenum, record_filenode, record_forknum, record_blocknum)));
buf = ReadBufferExtended(rel, record_forknum, record_blocknum, RBM_NORMAL, NULL);
ReleaseBuffer(buf);
++blocks_restored;
}
}
break;
case 'N':
{
BlockNumber block;
Assert(record_blocknum != InvalidBlockNumber);
if (record_blocknum == InvalidBlockNumber)
ereport(ERROR,
(errmsg("found a block range record without a preceeding block record")));
fileRead(&record_range, sizeof(int), file, false, filepath);
if (skip_relation || skip_fork || skip_block)
continue;
ereport(log_level,
(errmsg("reader %d reading range filenode %u forknum %d blocknum %u range %u",
filenum, record_filenode, record_forknum, record_blocknum, record_range)));
for (block = record_blocknum + 1; block <= (record_blocknum + record_range); ++block)
{
Buffer buf;
/*
* Don't try to read past the file; the file may have been
* shrunk by a vaccum operation.
*/
if (block >= nblocks)
{
ereport(log_level,
(errmsg("reader %d skipping block range filenode %u forknum %d start %u end %u",
filenum, record_filenode, record_forknum,
block, record_blocknum + record_range)));
break;
}
buf = ReadBufferExtended(rel, record_forknum, block, RBM_NORMAL, NULL);
ReleaseBuffer(buf);
++blocks_restored;
}
}
break;
default:
{
ereport(ERROR,
(errmsg("found unexpected save-file marker %x - %c)", record_type, record_type)));
Assert(false);
}
break;
}
}
if (rel)
relation_close(rel, AccessShareLock);
ereport(LOG,
(errmsg("Block Reader %d: restored %u blocks",
filenum, blocks_restored)));
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
fileClose(file, filepath);
/* Remove the save-file */
if (remove(filepath) != 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("error removing file \"%s\" : %m", filepath)));
}
/* --------------------------------
* InitPostgres
* Initialize POSTGRES.
*
* Note:
* Be very careful with the order of calls in the InitPostgres function.
* --------------------------------
*/
void
InitPostgres(const char *dbname, const char *username)
{
bool bootstrap = IsBootstrapProcessingMode();
/*
* Set up the global variables holding database id and path.
*
* We take a shortcut in the bootstrap case, otherwise we have to look up
* the db name in pg_database.
*/
if (bootstrap)
{
MyDatabaseId = TemplateDbOid;
SetDatabasePath(GetDatabasePath(MyDatabaseId));
}
else
{
char *fullpath,
datpath[MAXPGPATH];
/*
* Formerly we validated DataDir here, but now that's done
* earlier.
*/
/*
* Find oid and path of the database we're about to open. Since
* we're not yet up and running we have to use the hackish
* GetRawDatabaseInfo.
*/
GetRawDatabaseInfo(dbname, &MyDatabaseId, datpath);
if (!OidIsValid(MyDatabaseId))
ereport(FATAL,
(errcode(ERRCODE_UNDEFINED_DATABASE),
errmsg("database \"%s\" does not exist",
dbname)));
fullpath = GetDatabasePath(MyDatabaseId);
/* Verify the database path */
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);
if (chdir(fullpath) == -1)
ereport(FATAL,
(errcode_for_file_access(),
errmsg("could not change directory to \"%s\": %m",
fullpath)));
SetDatabasePath(fullpath);
}
/*
* Code after this point assumes we are in the proper directory!
*/
/*
* Set up my per-backend PGPROC struct in shared memory. (We need
* to know MyDatabaseId before we can do this, since it's entered into
* the PGPROC struct.)
*/
InitProcess();
/*
* Initialize my entry in the shared-invalidation manager's array of
* per-backend data. (Formerly this came before InitProcess, but now
* it must happen after, because it uses MyProc.) Once I have done
* this, I am visible to other backends!
*
* Sets up MyBackendId, a unique backend identifier.
*/
MyBackendId = InvalidBackendId;
InitBackendSharedInvalidationState();
if (MyBackendId > MaxBackends || MyBackendId <= 0)
elog(FATAL, "bad backend id: %d", MyBackendId);
/*
* Initialize the transaction system override state.
*/
AmiTransactionOverride(bootstrap);
/*
* Initialize the relation descriptor cache. This must create at
* least the minimum set of "nailed-in" cache entries. No catalog
* access happens here.
*/
RelationCacheInitialize();
/*
* Initialize all the system catalog caches. Note that no catalog
* access happens here; we only set up the cache structure.
*/
InitCatalogCache();
/* Initialize portal manager */
EnablePortalManager();
/*
* Initialize the deferred trigger manager --- must happen before
* first transaction start.
*/
DeferredTriggerInit();
/* start a new transaction here before access to db */
if (!bootstrap)
StartTransactionCommand();
/*
* It's now possible to do real access to the system catalogs.
*
* Replace faked-up relcache entries with correct info.
*/
RelationCacheInitializePhase2();
/*
* Figure out our postgres user id. In standalone mode we use a fixed
* id, otherwise we figure it out from the authenticated user name.
*/
if (bootstrap)
InitializeSessionUserIdStandalone();
else if (!IsUnderPostmaster)
{
InitializeSessionUserIdStandalone();
if (!ThereIsAtLeastOneUser())
ereport(WARNING,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("no users are defined in this database system"),
errhint("You should immediately run CREATE USER \"%s\" WITH SYSID %d CREATEUSER;.",
username, BOOTSTRAP_USESYSID)));
}
else
{
/* normal multiuser case */
InitializeSessionUserId(username);
}
/*
* Unless we are bootstrapping, double-check that InitMyDatabaseInfo()
* got a correct result. We can't do this until all the
* database-access infrastructure is up.
*/
if (!bootstrap)
ReverifyMyDatabase(dbname);
/*
* Final phase of relation cache startup: write a new cache file if
* necessary. This is done after ReverifyMyDatabase to avoid writing
* a cache file into a dead database.
*/
RelationCacheInitializePhase3();
/*
* Check a normal user hasn't connected to a superuser reserved slot.
* We can't do this till after we've read the user information, and we
* must do it inside a transaction since checking superuserness may
* require database access. The superuser check is probably the most
* expensive part; don't do it until necessary.
*/
if (ReservedBackends > 0 &&
CountEmptyBackendSlots() < ReservedBackends &&
!superuser())
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 done ReverifyMyDatabase.
*/
/* set default namespace search path */
InitializeSearchPath();
/* initialize client encoding */
InitializeClientEncoding();
/*
* Now all default states are fully set up. Report them to client if
* appropriate.
*/
BeginReportingGUCOptions();
/*
* Set up process-exit callback to do pre-shutdown cleanup. This
* should be last because we want shmem_exit to call this routine
* before the exit callbacks that are registered by buffer manager,
* lock manager, etc. We need to run this code before we close down
* database access!
*/
on_shmem_exit(ShutdownPostgres, 0);
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
}
static void
worker_spi_main(Datum main_arg)
{
/* Register functions for SIGTERM/SIGHUP management */
pqsignal(SIGHUP, worker_spi_sighup);
pqsignal(SIGTERM, worker_spi_sigterm);
/* We're now ready to receive signals */
BackgroundWorkerUnblockSignals();
/* Connect to our database */
BackgroundWorkerInitializeConnection("postgres", NULL);
while (!got_sigterm)
{
int ret;
int rc;
StringInfoData buf;
/*
* Background workers mustn't call usleep() or any direct equivalent:
* instead, they may wait on their process latch, which sleeps as
* necessary, but is awakened if postmaster dies. That way the
* background process goes away immediately in an emergency.
*/
rc = WaitLatch(&MyProc->procLatch,
WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
1000L);
ResetLatch(&MyProc->procLatch);
/* emergency bailout if postmaster has died */
if (rc & WL_POSTMASTER_DEATH)
proc_exit(1);
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
initStringInfo(&buf);
/* Build the query string */
appendStringInfo(&buf,
"SELECT count(*) FROM pg_class;");
ret = SPI_execute(buf.data, true, 0);
/* Some error messages in case of incorrect handling */
if (ret != SPI_OK_SELECT)
elog(FATAL, "SPI_execute failed: error code %d", ret);
if (SPI_processed > 0)
{
int32 count;
bool isnull;
count = DatumGetInt32(SPI_getbinval(SPI_tuptable->vals[0],
SPI_tuptable->tupdesc,
1, &isnull));
elog(LOG, "Currently %d relations in database",
count);
}
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
}
proc_exit(0);
}
/*
* This function receives a command names and a query and it produces a protocol valid command string
* which can be send to the server based on the format of the result set. It handles memory contexts
* appropriately, returning a pointer to the command string in the current context when called.
*
*/
char* exec_to_command(const char* command, char* q) {
StringInfoData resultbuf;
char* result;
int i, j, processed, retval;
SPITupleTable *coltuptable;
MemoryContext pre_context;
MemoryContext spi_conn_context;
// elog(LOG, "%s", command); //prints the current command running
pre_context = CurrentMemoryContext;
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
retval = SPI_execute(q, false, 0);
if (retval != SPI_OK_SELECT) {
elog(LOG, "Database SELECT execution failed: %d", retval);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
result = pstrdup("");
return result;
}
processed = SPI_processed;
coltuptable = SPI_tuptable;
initStringInfo(&resultbuf);
appendStringInfoString(&resultbuf, command);
appendStringInfoString(&resultbuf, ";"); //artisinal semicolon
if (coltuptable != NULL) {
for(i = 0; i < processed; i++) {
for(j = 1; j <= coltuptable->tupdesc->natts; j++) {
if (SPI_getvalue(coltuptable->vals[i], coltuptable->tupdesc, j) != NULL) {
appendStringInfoString(&resultbuf, SPI_getvalue(coltuptable->vals[i], coltuptable->tupdesc, j));
}
appendStringInfo(&resultbuf, FIELD_DELIMIT);
}
appendStringInfo(&resultbuf,REC_DELIMIT);
}
}
appendStringInfo(&resultbuf, CDELIMIT);
spi_conn_context = MemoryContextSwitchTo(pre_context);
result = pstrdup(resultbuf.data);
MemoryContextSwitchTo(spi_conn_context);
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
return result;
}
/*
* update segment configuration in catalog and shared memory
*/
static bool
probeUpdateConfig(FtsSegmentStatusChange *changes, int changeCount)
{
Relation configrel;
Relation histrel;
SysScanDesc sscan;
ScanKeyData scankey;
HeapTuple configtuple;
HeapTuple newtuple;
HeapTuple histtuple;
Datum configvals[Natts_gp_segment_configuration];
bool confignulls[Natts_gp_segment_configuration] = { false };
bool repls[Natts_gp_segment_configuration] = { false };
Datum histvals[Natts_gp_configuration_history];
bool histnulls[Natts_gp_configuration_history] = { false };
bool valid;
bool primary;
bool changelogging;
int i;
char desc[SQL_CMD_BUF_SIZE];
/*
* Commit/abort transaction below will destroy
* CurrentResourceOwner. We need it for catalog reads.
*/
ResourceOwner save = CurrentResourceOwner;
StartTransactionCommand();
elog(LOG, "probeUpdateConfig called for %d changes", changeCount);
histrel = heap_open(GpConfigHistoryRelationId,
RowExclusiveLock);
configrel = heap_open(GpSegmentConfigRelationId,
RowExclusiveLock);
for (i = 0; i < changeCount; i++)
{
FtsSegmentStatusChange *change = &changes[i];
valid = (changes[i].newStatus & FTS_STATUS_ALIVE);
primary = (changes[i].newStatus & FTS_STATUS_PRIMARY);
changelogging = (changes[i].newStatus & FTS_STATUS_CHANGELOGGING);
if (changelogging)
{
Assert(failover_strategy == 'f');
Assert(primary && valid);
}
Assert((valid || !primary) && "Primary cannot be down");
/*
* Insert new tuple into gp_configuration_history catalog.
*/
histvals[Anum_gp_configuration_history_time-1] =
TimestampTzGetDatum(GetCurrentTimestamp());
histvals[Anum_gp_configuration_history_dbid-1] =
Int16GetDatum(changes[i].dbid);
snprintf(desc, sizeof(desc),
"FTS: content %d fault marking status %s%s role %c",
change->segindex, valid ? "UP" : "DOWN",
(changelogging) ? " mode: change-tracking" : "",
primary ? 'p' : 'm');
histvals[Anum_gp_configuration_history_desc-1] =
CStringGetTextDatum(desc);
histtuple = heap_form_tuple(RelationGetDescr(histrel), histvals, histnulls);
simple_heap_insert(histrel, histtuple);
CatalogUpdateIndexes(histrel, histtuple);
/*
* Find and update gp_segment_configuration tuple.
*/
ScanKeyInit(&scankey,
Anum_gp_segment_configuration_dbid,
BTEqualStrategyNumber, F_INT2EQ,
Int16GetDatum(changes[i].dbid));
sscan = systable_beginscan(configrel, GpSegmentConfigDbidIndexId,
true, SnapshotNow, 1, &scankey);
configtuple = systable_getnext(sscan);
if (!HeapTupleIsValid(configtuple))
{
elog(ERROR, "FTS cannot find dbid=%d in %s", changes[i].dbid,
RelationGetRelationName(configrel));
}
configvals[Anum_gp_segment_configuration_role-1] =
CharGetDatum(primary ? 'p' : 'm');
repls[Anum_gp_segment_configuration_role-1] = true;
configvals[Anum_gp_segment_configuration_status-1] =
CharGetDatum(valid ? 'u' : 'd');
repls[Anum_gp_segment_configuration_status-1] = true;
if (changelogging)
{
configvals[Anum_gp_segment_configuration_mode-1] =
CharGetDatum('c');
}
repls[Anum_gp_segment_configuration_mode-1] = changelogging;
newtuple = heap_modify_tuple(configtuple, RelationGetDescr(configrel),
configvals, confignulls, repls);
simple_heap_update(configrel, &configtuple->t_self, newtuple);
CatalogUpdateIndexes(configrel, newtuple);
systable_endscan(sscan);
pfree(newtuple);
/*
* Update shared memory
*/
ftsProbeInfo->fts_status[changes[i].dbid] = changes[i].newStatus;
}
heap_close(histrel, RowExclusiveLock);
heap_close(configrel, RowExclusiveLock);
SIMPLE_FAULT_INJECTOR(FtsWaitForShutdown);
/*
* Do not block shutdown. We will always get a change to update
* gp_segment_configuration in subsequent probes upon database
* restart.
*/
if (shutdown_requested)
{
elog(LOG, "Shutdown in progress, ignoring FTS prober updates.");
return false;
}
CommitTransactionCommand();
CurrentResourceOwner = save;
return true;
}
/**
* Marks the given db as in-sync in the segment configuration.
*/
void
FtsMarkSegmentsInSync(CdbComponentDatabaseInfo *primary, CdbComponentDatabaseInfo *mirror)
{
if (!FTS_STATUS_ISALIVE(primary->dbid, ftsProbeInfo->fts_status) ||
!FTS_STATUS_ISALIVE(mirror->dbid, ftsProbeInfo->fts_status) ||
!FTS_STATUS_ISPRIMARY(primary->dbid, ftsProbeInfo->fts_status) ||
FTS_STATUS_ISPRIMARY(mirror->dbid, ftsProbeInfo->fts_status) ||
FTS_STATUS_IS_SYNCED(primary->dbid, ftsProbeInfo->fts_status) ||
FTS_STATUS_IS_SYNCED(mirror->dbid, ftsProbeInfo->fts_status) ||
FTS_STATUS_IS_CHANGELOGGING(primary->dbid, ftsProbeInfo->fts_status) ||
FTS_STATUS_IS_CHANGELOGGING(mirror->dbid, ftsProbeInfo->fts_status))
{
FtsRequestPostmasterShutdown(primary, mirror);
}
if (ftsProbeInfo->fts_pauseProbes)
{
return;
}
uint8 segStatus=0;
Relation configrel;
Relation histrel;
ScanKeyData scankey;
SysScanDesc sscan;
HeapTuple configtuple;
HeapTuple newtuple;
HeapTuple histtuple;
Datum configvals[Natts_gp_segment_configuration];
bool confignulls[Natts_gp_segment_configuration] = { false };
bool repls[Natts_gp_segment_configuration] = { false };
Datum histvals[Natts_gp_configuration_history];
bool histnulls[Natts_gp_configuration_history] = { false };
char *desc = "FTS: changed segment to insync from resync.";
/*
* Commit/abort transaction below will destroy
* CurrentResourceOwner. We need it for catalog reads.
*/
ResourceOwner save = CurrentResourceOwner;
StartTransactionCommand();
/* update primary */
segStatus = ftsProbeInfo->fts_status[primary->dbid];
segStatus |= FTS_STATUS_SYNCHRONIZED;
ftsProbeInfo->fts_status[primary->dbid] = segStatus;
/* update mirror */
segStatus = ftsProbeInfo->fts_status[mirror->dbid];
segStatus |= FTS_STATUS_SYNCHRONIZED;
ftsProbeInfo->fts_status[mirror->dbid] = segStatus;
histrel = heap_open(GpConfigHistoryRelationId,
RowExclusiveLock);
configrel = heap_open(GpSegmentConfigRelationId,
RowExclusiveLock);
/* update gp_segment_configuration to insync */
ScanKeyInit(&scankey,
Anum_gp_segment_configuration_dbid,
BTEqualStrategyNumber, F_INT2EQ,
Int16GetDatum(primary->dbid));
sscan = systable_beginscan(configrel, GpSegmentConfigDbidIndexId,
true, SnapshotNow, 1, &scankey);
configtuple = systable_getnext(sscan);
if (!HeapTupleIsValid(configtuple))
{
elog(ERROR,"FTS cannot find dbid (%d, %d) in %s", primary->dbid,
mirror->dbid, RelationGetRelationName(configrel));
}
configvals[Anum_gp_segment_configuration_mode-1] = CharGetDatum('s');
repls[Anum_gp_segment_configuration_mode-1] = true;
newtuple = heap_modify_tuple(configtuple, RelationGetDescr(configrel),
configvals, confignulls, repls);
simple_heap_update(configrel, &configtuple->t_self, newtuple);
CatalogUpdateIndexes(configrel, newtuple);
systable_endscan(sscan);
ScanKeyInit(&scankey,
Anum_gp_segment_configuration_dbid,
BTEqualStrategyNumber, F_INT2EQ,
Int16GetDatum(mirror->dbid));
sscan = systable_beginscan(configrel, GpSegmentConfigDbidIndexId,
true, SnapshotNow, 1, &scankey);
configtuple = systable_getnext(sscan);
if (!HeapTupleIsValid(configtuple))
{
elog(ERROR,"FTS cannot find dbid (%d, %d) in %s", primary->dbid,
mirror->dbid, RelationGetRelationName(configrel));
}
newtuple = heap_modify_tuple(configtuple, RelationGetDescr(configrel),
configvals, confignulls, repls);
simple_heap_update(configrel, &configtuple->t_self, newtuple);
CatalogUpdateIndexes(configrel, newtuple);
systable_endscan(sscan);
/* update configuration history */
histvals[Anum_gp_configuration_history_time-1] =
TimestampTzGetDatum(GetCurrentTimestamp());
histvals[Anum_gp_configuration_history_dbid-1] =
Int16GetDatum(primary->dbid);
histvals[Anum_gp_configuration_history_desc-1] =
CStringGetTextDatum(desc);
histtuple = heap_form_tuple(RelationGetDescr(histrel), histvals, histnulls);
simple_heap_insert(histrel, histtuple);
CatalogUpdateIndexes(histrel, histtuple);
histvals[Anum_gp_configuration_history_dbid-1] =
Int16GetDatum(mirror->dbid);
histtuple = heap_form_tuple(RelationGetDescr(histrel), histvals, histnulls);
simple_heap_insert(histrel, histtuple);
CatalogUpdateIndexes(histrel, histtuple);
ereport(LOG,
(errmsg("FTS: resynchronization of mirror (dbid=%d, content=%d) on %s:%d has completed.",
mirror->dbid, mirror->segindex, mirror->address, mirror->port ),
errSendAlert(true)));
heap_close(histrel, RowExclusiveLock);
heap_close(configrel, RowExclusiveLock);
/*
* Do not block shutdown. We will always get a change to update
* gp_segment_configuration in subsequent probes upon database
* restart.
*/
if (shutdown_requested)
{
elog(LOG, "Shutdown in progress, ignoring FTS prober updates.");
return;
}
CommitTransactionCommand();
CurrentResourceOwner = save;
}
