/*
* Verify that a freshly-read page looks sane.
*/
void
gistcheckpage(Relation rel, Buffer buf)
{
Page page = BufferGetPage(buf);
/*
* ReadBuffer verifies that every newly-read page passes
* PageHeaderIsValid, which means it either contains a reasonably sane
* page header or is all-zero. We have to defend against the all-zero
* case, however.
*/
if (PageIsNew(page))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" contains unexpected zero page at block %u",
RelationGetRelationName(rel),
BufferGetBlockNumber(buf)),
errhint("Please REINDEX it."),
errSendAlert(true)));
/*
* Additionally check that the special area looks sane.
*/
if (((PageHeader) (page))->pd_special !=
(BLCKSZ - MAXALIGN(sizeof(GISTPageOpaqueData))))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" contains corrupted page at block %u",
RelationGetRelationName(rel),
BufferGetBlockNumber(buf)),
errhint("Please REINDEX it."),
errSendAlert(true)));
}
int
FileRepConnServer_StartListener(
char *hostAddress,
int portLocal)
{
int status = STATUS_OK;
int i;
for (i=0; i < FILEREP_MAX_LISTEN; i++) {
listenSocket[i] = -1;
}
/* NOTE check if family AF_UNIX has to be considered as well */
status = StreamServerPort(
AF_UNSPEC,
hostAddress,
(unsigned short) portLocal,
NULL,
listenSocket,
FILEREP_MAX_LISTEN);
if (status != STATUS_OK) {
ereport(WARNING,
(errcode_for_socket_access(),
errmsg("could not start listener, host:'%s' port:'%d': %m",
hostAddress,
portLocal),
errSendAlert(true),
FileRep_errcontext()));
}
return status;
}
void
FtsHandleNetFailure(SegmentDatabaseDescriptor ** segDB, int numOfFailed)
{
elog(LOG, "FtsHandleNetFailure: numOfFailed %d", numOfFailed);
ereport(ERROR, (errmsg_internal("MPP detected %d segment failures, system is reconnected", numOfFailed),
errSendAlert(true)));
}
int
FileRepConnClient_EstablishConnection(
char *hostAddress,
int port,
bool reportError)
{
int status = STATUS_OK;
char portbuf[11];
char timeoutbuf[11];
const char *keys[5];
const char *vals[5];
/* FileRepConnClient_CloseConnection();*/
snprintf(portbuf, sizeof(portbuf), "%d", port);
snprintf(timeoutbuf, sizeof(timeoutbuf), "%d", gp_segment_connect_timeout);
keys[0] = "host";
vals[0] = hostAddress;
keys[1] = "port";
vals[1] = portbuf;
keys[2] = "dbname";
vals[2] = "postgres";
keys[3] = "connect_timeout";
vals[3] = timeoutbuf;
keys[4] = NULL;
vals[4] = NULL;
filerep_conn = PQconnectdbParams(keys, vals, false);
if (PQstatus(filerep_conn) != CONNECTION_OK)
{
if (reportError || Debug_filerep_print)
ereport(WARNING,
(errcode_for_socket_access(),
errmsg("could not establish connection with server, host:'%s' port:'%d' err:'%s' : %m",
hostAddress,
port,
PQerrorMessage(filerep_conn)),
errSendAlert(true),
FileRep_errcontext()));
status = STATUS_ERROR;
if (filerep_conn)
{
PQfinish(filerep_conn);
filerep_conn = NULL;
}
}
/* NOTE Handle error message see ftsprobe.c */
return status;
}
/*
* Re-Configure the system: if someone has noticed that the status
* version has been updated, they call this to verify that they've got
* the right configuration.
*
* NOTE: This *always* destroys gangs. And also attempts to inform the
* fault-prober to do a full scan.
*/
void
FtsReConfigureMPP(bool create_new_gangs)
{
/* need to scan to pick up the latest view */
detectFailedConnections();
local_fts_statusVersion = ftsProbeInfo->fts_statusVersion;
ereport(LOG, (errmsg_internal("FTS: reconfiguration is in progress"),
errSendAlert(true)));
disconnectAndDestroyAllGangs();
/* Caller should throw an error. */
return;
}
/*
* ProcessRoleGUC --
* We now process pg_authid.rolconfig separately from InitializeSessionUserId,
* since it's too early to access toast table before initializing all
* relcaches in phase3.
*/
static void
ProcessRoleGUC(void)
{
cqContext *pcqCtx;
Oid roleId;
HeapTuple roleTup;
Datum datum;
bool isnull;
/* This should have been set by now */
roleId = GetUserId();
Assert(OidIsValid(roleId));
pcqCtx = caql_beginscan(
NULL,
cql("SELECT * FROM pg_authid "
" WHERE oid = :1",
ObjectIdGetDatum(roleId)));
roleTup = caql_getnext(pcqCtx);
if (!HeapTupleIsValid(roleTup))
ereport(FATAL,
(errcode(ERRCODE_INVALID_AUTHORIZATION_SPECIFICATION),
errmsg("role %u does not exist", roleId), errSendAlert(false)));
/*
* Set up user-specific configuration variables. This is a good place to
* do it so we don't have to read pg_authid twice during session startup.
*/
datum = caql_getattr(pcqCtx,
Anum_pg_authid_rolconfig, &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_authid was checked when it was
* inserted.
*/
ProcessGUCArray(a, PGC_SUSET, PGC_S_USER, GUC_ACTION_SET);
}
caql_endscan(pcqCtx);
}
bool QDMirroringWriteCheck(void)
{
bool giveWarning = false;
QDMIRRORDisabledReason disabledReason = QDMIRROR_DISABLEDREASON_NONE;
struct timeval lastLogTimeVal = {0, 0};
if (ftsQDMirrorInfo == NULL)
return false; // Don't know yet.
LWLockAcquire(ftsQDMirrorLock, LW_EXCLUSIVE);
if (ftsQDMirrorInfo->state == QDMIRROR_STATE_SYNCHRONIZED)
{
LWLockRelease(ftsQDMirrorLock);
return true;
}
if (ftsQDMirrorInfo->QDMirroringNotSynchronizedWarningGiven == false &&
ftsQDMirrorInfo->state == QDMIRROR_STATE_DISABLED)
{
giveWarning = true;
ftsQDMirrorInfo->QDMirroringNotSynchronizedWarningGiven = true;
disabledReason = ftsQDMirrorInfo->disabledReason;
lastLogTimeVal = ftsQDMirrorInfo->lastLogTimeVal;
}
LWLockRelease(ftsQDMirrorLock);
if (giveWarning)
{
char logTimeStr[100];
QDMirroringFormatTime(logTimeStr, sizeof(logTimeStr),
&lastLogTimeVal);
ereport(LOG,
(errcode(ERRCODE_GP_INTERCONNECTION_ERROR),
errmsg("Master mirroring is not synchronized as of %s (%s). "
"The GPDB system is currently not highly-available",
logTimeStr,
QDMirroringDisabledReasonToString(disabledReason)),
errSendAlert(true)));
}
return false;
}
/*
* FtsHandleGangConnectionFailure is called by createGang during
* creating connections return true if error need to be thrown
*/
bool
FtsHandleGangConnectionFailure(SegmentDatabaseDescriptor * segdbDesc, int size)
{
int i;
bool dtx_active;
bool reportError = false;
bool realFaultFound = false;
bool forceRescan=true;
for (i = 0; i < size; i++)
{
if (PQstatus(segdbDesc[i].conn) != CONNECTION_OK)
{
CdbComponentDatabaseInfo *segInfo = segdbDesc[i].segment_database_info;
elog(DEBUG2, "FtsHandleGangConnectionFailure: looking for real fault on segment dbid %d", segInfo->dbid);
if (!FtsTestConnection(segInfo, forceRescan))
{
elog(DEBUG2, "found fault with segment dbid %d", segInfo->dbid);
realFaultFound = true;
/* that at least one fault exists is enough, for now */
break;
}
forceRescan = false; /* only force the rescan on the first call. */
}
}
if (!realFaultFound)
{
/* If we successfully tested the gang and didn't notice a
* failure, our caller must've seen some kind of transient
* failure when the gang was originally constructed ... */
elog(DEBUG2, "FtsHandleGangConnectionFailure: no real fault found!");
return false;
}
if (!isFTSEnabled())
{
return false;
}
ereport(LOG, (errmsg_internal("FTS: reconfiguration is in progress")));
forceRescan = true;
for (i = 0; i < size; i++)
{
CdbComponentDatabaseInfo *segInfo = segdbDesc[i].segment_database_info;
if (PQstatus(segdbDesc[i].conn) != CONNECTION_OK)
{
if (!FtsTestConnection(segInfo, forceRescan))
{
ereport(LOG, (errmsg_internal("FTS: found bad segment with dbid %d", segInfo->dbid),
errSendAlert(true)));
/* probe process has already marked segment down. */
}
forceRescan = false; /* only force rescan on first call. */
}
}
if (gangsExist())
{
reportError = true;
disconnectAndDestroyAllGangs();
}
/*
* KLUDGE: Do not error out if we are attempting a DTM protocol retry
*/
if (DistributedTransactionContext == DTX_CONTEXT_QD_RETRY_PHASE_2)
{
return false;
}
/* is there a transaction active ? */
dtx_active = isCurrentDtxActive();
/* When the error is raised, it will abort the current DTM transaction */
if (dtx_active)
{
elog((Debug_print_full_dtm ? LOG : DEBUG5),
"FtsHandleGangConnectionFailure found an active DTM transaction (returning true).");
return true;
}
/*
* error out if this sets read only flag, at this stage the read only
* transaction checking has passed, so error out, but do not error out if
* tm is in recovery
*/
if ((*ftsReadOnlyFlag && !isTMInRecovery()) || reportError)
return true;
elog((Debug_print_full_dtm ? LOG : DEBUG5),
"FtsHandleGangConnectionFailure returning false.");
return false;
}
/* --------------------------------
* 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 = IsAutoVacuumProcess();
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),
errSendAlert(false)));
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),
errOmitLocation(true),
errSendAlert(false)));
/* 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();
// get temporary directory for QD
if (!bootstrap && Gp_role == GP_ROLE_DISPATCH)
{
if (get_tmpdir_from_rm)
{
getLocalTmpDirFromMasterRM();
}
else
{
getLocalTmpDirFromMasterConfig(gp_session_id);
elog(LOG, "getLocalTmpDirFromMasterConfig session_id:%d tmpdir:%s", gp_session_id, LocalTempPath);
}
}
/* Initialize SessionState entry */
SessionState_Init();
/* Initialize memory protection */
GPMemoryProtect_Init();
/*
* 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();
/*
* 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)
{
if (MyDatabaseId == TemplateDbOid)
LockSharedObject(DatabaseRelationId, MyDatabaseId, 0,
AccessShareLock);
else
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);
/*
* 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"),
errOmitLocation(true),
errSendAlert(true)));
/*
* 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();
/*
* 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 );
}
/* close the transaction we started above */
if (!bootstrap)
CommitTransactionCommand();
return am_superuser;
}
/*
* 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 real pg_database row */
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 && !IsAutoVacuumProcess())
{
/*
* Check that the database is currently allowing connections.
* (exception during upgrade_mode)
*/
if (gp_upgrade_mode && !dbform->datallowconn)
elog(INFO, "Connecting to no-connection db in upgrade mode.");
if (!dbform->datallowconn && !gp_upgrade_mode)
ereport(FATAL,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("database \"%s\" is not currently accepting connections",
name),
errOmitLocation(true),
errSendAlert(false)));
/*
* 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."),
errOmitLocation(true),
errSendAlert(false)));
/*
* 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),
errOmitLocation(true),
errSendAlert(false)));
}
/*
* OK, we're golden. Next to-do item is to save the encoding info out of
* the pg_database tuple.
*/
if (GpIdentity.segindex == UNINITIALIZED_GP_IDENTITY_VALUE ||
GpIdentity.segindex == MASTER_CONTENT_ID)
SetDatabaseEncoding(dbform->encoding);
else
SetDatabaseEncoding(MyProcPort->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);
ProcessGUCArray(a, PGC_S_DATABASE);
}
}
ReleaseSysCache(tup);
}
/*
*
* FileRepAckPrimary_StartReceiver
*/
void
FileRepAckPrimary_StartReceiver(void)
{
int status = STATUS_OK;
struct timeval currentTime;
pg_time_t beginTime = 0;
pg_time_t endTime = 0;
int retval = 0;
FileRep_InsertConfigLogEntry("start receiver ack");
{
char tmpBuf[FILEREP_MAX_LOG_DESCRIPTION_LEN];
snprintf(tmpBuf, sizeof(tmpBuf), "primary address(port) '%s(%d)' mirror address(port) '%s(%d)' ",
fileRepPrimaryHostAddress,
fileRepPrimaryPort,
fileRepMirrorHostAddress,
fileRepMirrorPort);
FileRep_InsertConfigLogEntry(tmpBuf);
}
FileRepAckPrimary_ShmemReInit();
Insist(fileRepRole == FileRepPrimaryRole);
if (filerep_inject_listener_fault)
{
status = STATUS_ERROR;
ereport(WARNING,
(errmsg("mirror failure, "
"injected fault by guc filerep_inject_listener_fault, "
"failover requested"),
FileRep_errcontext()));
FileRep_SetSegmentState(SegmentStateFault, FaultTypeMirror);
FileRepSubProcess_SetState(FileRepStateFault);
FileRepSubProcess_ProcessSignals();
return;
}
status = FileRepConnServer_StartListener(
fileRepPrimaryHostAddress,
fileRepPrimaryPort);
gettimeofday(¤tTime, NULL);
beginTime = (pg_time_t) currentTime.tv_sec;
while (1) {
if (status != STATUS_OK)
{
FileRep_SetSegmentState(SegmentStateFault, FaultTypeMirror);
FileRepSubProcess_SetState(FileRepStateFault);
}
while (FileRepSubProcess_GetState() == FileRepStateFault) {
FileRepSubProcess_ProcessSignals();
pg_usleep(50000L); /* 50 ms */
}
if (FileRepSubProcess_GetState() == FileRepStateShutdown) {
break;
}
PG_SETMASK(&BlockSig);
retval = FileRepConnServer_Select();
PG_SETMASK(&UnBlockSig);
gettimeofday(¤tTime, NULL);
endTime = (pg_time_t) currentTime.tv_sec;
if ((endTime - beginTime) > gp_segment_connect_timeout)
{
ereport(WARNING,
(errmsg("mirror failure, "
"no connection was established from client from mirror, "
"primary address(port) '%s(%d)' mirror address(port) '%s(%d)' timeout reached '%d' "
"failover requested",
fileRepPrimaryHostAddress,
fileRepPrimaryPort,
fileRepMirrorHostAddress,
fileRepMirrorPort,
gp_segment_connect_timeout),
errSendAlert(true),
FileRep_errcontext()));
status = STATUS_ERROR;
continue;
}
/*
* check and process any signals received
* The routine returns TRUE if the received signal requests
* process shutdown.
*/
if (FileRepSubProcess_ProcessSignals()) {
continue;
}
if (retval < 0) {
status = STATUS_ERROR;
continue;
}
if (retval == 0) {
continue;
}
Assert(retval > 0);
status = FileRepConnServer_CreateConnection();
if (status != STATUS_OK) {
continue;
}
status = FileRepConnServer_ReceiveStartupPacket();
if (status != STATUS_OK) {
continue;
}
fileRepShmemArray[0]->state = FileRepStateInitialization;
status = FileRepAckPrimary_RunReceiver();
} // while(1)
FileRepConnServer_CloseConnection();
return;
}
/*
* PageIndexMultiDelete
*
* This routine handles the case of deleting multiple tuples from an
* index page at once. It is considerably faster than a loop around
* PageIndexTupleDelete ... however, the caller *must* supply the array
* of item numbers to be deleted in item number order!
*/
void
PageIndexMultiDelete(Page page, OffsetNumber *itemnos, int nitems)
{
PageHeader phdr = (PageHeader) page;
Offset pd_lower = phdr->pd_lower;
Offset pd_upper = phdr->pd_upper;
Offset pd_special = phdr->pd_special;
itemIdSort itemidbase,
itemidptr;
ItemId lp;
int nline,
nused;
int i;
Size totallen;
Offset upper;
Size size;
unsigned offset;
int nextitm;
OffsetNumber offnum;
/*
* If there aren't very many items to delete, then retail
* PageIndexTupleDelete is the best way. Delete the items in reverse
* order so we don't have to think about adjusting item numbers for
* previous deletions.
*
* TODO: tune the magic number here
*/
if (nitems <= 2)
{
while (--nitems >= 0)
PageIndexTupleDelete(page, itemnos[nitems]);
return;
}
/*
* As with PageRepairFragmentation, paranoia seems justified.
*/
if (pd_lower < SizeOfPageHeaderData ||
pd_lower > pd_upper ||
pd_upper > pd_special ||
pd_special > BLCKSZ ||
pd_special != MAXALIGN(pd_special))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted page pointers: lower = %u, upper = %u, special = %u",
pd_lower, pd_upper, pd_special),
errSendAlert(true)));
/*
* Scan the item pointer array and build a list of just the ones we are
* going to keep. Notice we do not modify the page yet, since we are
* still validity-checking.
*/
nline = PageGetMaxOffsetNumber(page);
itemidbase = (itemIdSort) palloc(sizeof(itemIdSortData) * nline);
itemidptr = itemidbase;
totallen = 0;
nused = 0;
nextitm = 0;
for (offnum = 1; offnum <= nline; offnum++)
{
lp = PageGetItemId(page, offnum);
size = ItemIdGetLength(lp);
offset = ItemIdGetOffset(lp);
if (offset < pd_upper ||
(offset + size) > pd_special ||
offset != MAXALIGN(offset))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item pointer: offset = %u, size = %u",
offset, (unsigned int) size),
errSendAlert(true)));
if (nextitm < nitems && offnum == itemnos[nextitm])
{
/* skip item to be deleted */
nextitm++;
}
else
{
itemidptr->offsetindex = nused; /* where it will go */
itemidptr->itemoff = offset;
itemidptr->olditemid = *lp;
itemidptr->alignedlen = MAXALIGN(size);
totallen += itemidptr->alignedlen;
itemidptr++;
nused++;
}
}
/* this will catch invalid or out-of-order itemnos[] */
if (nextitm != nitems)
elog(ERROR, "incorrect index offsets supplied");
if (totallen > (Size) (pd_special - pd_lower))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item lengths: total %u, available space %u",
(unsigned int) totallen, pd_special - pd_lower),
errSendAlert(true)));
/* sort itemIdSortData array into decreasing itemoff order */
qsort((char *) itemidbase, nused, sizeof(itemIdSortData),
itemoffcompare);
/* compactify page and install new itemids */
upper = pd_special;
for (i = 0, itemidptr = itemidbase; i < nused; i++, itemidptr++)
{
lp = PageGetItemId(page, itemidptr->offsetindex + 1);
upper -= itemidptr->alignedlen;
memmove((char *) page + upper,
(char *) page + itemidptr->itemoff,
itemidptr->alignedlen);
*lp = itemidptr->olditemid;
lp->lp_off = upper;
}
phdr->pd_lower = SizeOfPageHeaderData + nused * sizeof(ItemIdData);
phdr->pd_upper = upper;
pfree(itemidbase);
}
/*
* PageIndexTupleDelete
*
* This routine does the work of removing a tuple from an index page.
*
* Unlike heap pages, we compact out the line pointer for the removed tuple.
*/
void
PageIndexTupleDelete(Page page, OffsetNumber offnum)
{
PageHeader phdr = (PageHeader) page;
char *addr;
ItemId tup;
Size size;
unsigned offset;
int nbytes;
int offidx;
int nline;
/*
* As with PageRepairFragmentation, paranoia seems justified.
*/
if (phdr->pd_lower < SizeOfPageHeaderData ||
phdr->pd_lower > phdr->pd_upper ||
phdr->pd_upper > phdr->pd_special ||
phdr->pd_special > BLCKSZ)
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted page pointers: lower = %u, upper = %u, special = %u",
phdr->pd_lower, phdr->pd_upper, phdr->pd_special),
errSendAlert(true)));
nline = PageGetMaxOffsetNumber(page);
if ((int) offnum <= 0 || (int) offnum > nline)
elog(ERROR, "invalid index offnum: %u", offnum);
/* change offset number to offset index */
offidx = offnum - 1;
tup = PageGetItemId(page, offnum);
size = ItemIdGetLength(tup);
offset = ItemIdGetOffset(tup);
if (offset < phdr->pd_upper || (offset + size) > phdr->pd_special ||
offset != MAXALIGN(offset) || size != MAXALIGN(size))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item pointer: offset = %u, size = %u",
offset, (unsigned int) size),
errSendAlert(true)));
/*
* First, we want to get rid of the pd_linp entry for the index tuple. We
* copy all subsequent linp's back one slot in the array. We don't use
* PageGetItemId, because we are manipulating the _array_, not individual
* linp's.
*/
nbytes = phdr->pd_lower -
((char *) &phdr->pd_linp[offidx + 1] - (char *) phdr);
if (nbytes > 0)
memmove((char *) &(phdr->pd_linp[offidx]),
(char *) &(phdr->pd_linp[offidx + 1]),
nbytes);
/*
* Now move everything between the old upper bound (beginning of tuple
* space) and the beginning of the deleted tuple forward, so that space in
* the middle of the page is left free. If we've just deleted the tuple
* at the beginning of tuple space, then there's no need to do the copy
* (and bcopy on some architectures SEGV's if asked to move zero bytes).
*/
/* beginning of tuple space */
addr = (char *) page + phdr->pd_upper;
if (offset > phdr->pd_upper)
memmove(addr + size, addr, (int) (offset - phdr->pd_upper));
/* adjust free space boundary pointers */
phdr->pd_upper += size;
phdr->pd_lower -= sizeof(ItemIdData);
/*
* Finally, we need to adjust the linp entries that remain.
*
* Anything that used to be before the deleted tuple's data was moved
* forward by the size of the deleted tuple.
*/
if (!PageIsEmpty(page))
{
int i;
nline--; /* there's one less than when we started */
for (i = 1; i <= nline; i++)
{
ItemId ii = PageGetItemId(phdr, i);
if (ItemIdGetOffset(ii) <= offset)
ii->lp_off += size;
}
}
}
/*
* PageRepairFragmentation
*
* Frees fragmented space on a page.
* It doesn't remove unused line pointers! Please don't change this.
*
* This routine is usable for heap pages only, but see PageIndexMultiDelete.
*
* Returns number of unused line pointers on page. If "unused" is not NULL
* then the unused[] array is filled with indexes of unused line pointers.
*/
int
PageRepairFragmentation(Page page, OffsetNumber *unused)
{
Offset pd_lower = ((PageHeader) page)->pd_lower;
Offset pd_upper = ((PageHeader) page)->pd_upper;
Offset pd_special = ((PageHeader) page)->pd_special;
itemIdSort itemidbase,
itemidptr;
ItemId lp;
int nline,
nused;
int i;
Size totallen;
Offset upper;
/*
* It's worth the trouble to be more paranoid here than in most places,
* because we are about to reshuffle data in (what is usually) a shared
* disk buffer. If we aren't careful then corrupted pointers, lengths,
* etc could cause us to clobber adjacent disk buffers, spreading the data
* loss further. So, check everything.
*/
if (pd_lower < SizeOfPageHeaderData ||
pd_lower > pd_upper ||
pd_upper > pd_special ||
pd_special > BLCKSZ ||
pd_special != MAXALIGN(pd_special))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted page pointers: lower = %u, upper = %u, special = %u",
pd_lower, pd_upper, pd_special),
errSendAlert(true)));
nline = PageGetMaxOffsetNumber(page);
nused = 0;
for (i = 0; i < nline; i++)
{
lp = PageGetItemId(page, i + 1);
if (ItemIdDeleted(lp)) /* marked for deletion */
lp->lp_flags &= ~(LP_USED | LP_DELETE);
if (ItemIdIsUsed(lp))
nused++;
else if (unused)
unused[i - nused] = (OffsetNumber) i;
}
if (nused == 0)
{
/* Page is completely empty, so just reset it quickly */
for (i = 0; i < nline; i++)
{
lp = PageGetItemId(page, i + 1);
lp->lp_len = 0; /* indicate unused & deallocated */
}
((PageHeader) page)->pd_upper = pd_special;
}
else
{ /* nused != 0 */
/* Need to compact the page the hard way */
itemidbase = (itemIdSort) palloc(sizeof(itemIdSortData) * nused);
itemidptr = itemidbase;
totallen = 0;
for (i = 0; i < nline; i++)
{
lp = PageGetItemId(page, i + 1);
if (ItemIdIsUsed(lp))
{
itemidptr->offsetindex = i;
itemidptr->itemoff = ItemIdGetOffset(lp);
if (itemidptr->itemoff < (int) pd_upper ||
itemidptr->itemoff >= (int) pd_special)
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item pointer: %u",
itemidptr->itemoff),
errSendAlert(true)));
itemidptr->alignedlen = MAXALIGN(ItemIdGetLength(lp));
totallen += itemidptr->alignedlen;
itemidptr++;
}
else
{
lp->lp_len = 0; /* indicate unused & deallocated */
}
}
if (totallen > (Size) (pd_special - pd_lower))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item lengths: total %u, available space %u",
(unsigned int) totallen, pd_special - pd_lower),
errSendAlert(true)));
/* sort itemIdSortData array into decreasing itemoff order */
qsort((char *) itemidbase, nused, sizeof(itemIdSortData),
itemoffcompare);
/* compactify page */
upper = pd_special;
for (i = 0, itemidptr = itemidbase; i < nused; i++, itemidptr++)
{
lp = PageGetItemId(page, itemidptr->offsetindex + 1);
upper -= itemidptr->alignedlen;
memmove((char *) page + upper,
(char *) page + itemidptr->itemoff,
itemidptr->alignedlen);
lp->lp_off = upper;
}
((PageHeader) page)->pd_upper = upper;
pfree(itemidbase);
}
/* Set hint bit for PageAddItem */
if (nused < nline)
PageSetHasFreeLinePointers(page);
else
PageClearHasFreeLinePointers(page);
return (nline - nused);
}
/*
* _bitmap_init() -- initialize the bitmap index.
*
* Create the meta page, a new heap which stores the distinct values for
* the attributes to be indexed, a btree index on this new heap for searching
* those distinct values, and the first LOV page.
*/
void
_bitmap_init(Relation rel, Oid comptypeOid,
Oid heapOid, Oid indexOid,
Oid heapRelfilenode, Oid indexRelfilenode,
bool use_wal)
{
MIRROREDLOCK_BUFMGR_DECLARE;
BMMetaPage metapage;
Buffer metabuf;
Page page;
Buffer buf;
BMLOVItem lovItem;
OffsetNumber newOffset;
Page currLovPage;
OffsetNumber o;
/* sanity check */
if (RelationGetNumberOfBlocks(rel) != 0)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("cannot initialize non-empty bitmap index \"%s\"",
RelationGetRelationName(rel)),
errSendAlert(true)));
// -------- MirroredLock ----------
MIRROREDLOCK_BUFMGR_LOCK;
/* create the metapage */
metabuf = _bitmap_getbuf(rel, P_NEW, BM_WRITE);
page = BufferGetPage(metabuf);
Assert(PageIsNew(page));
/* initialize the LOV metadata */
_bitmap_create_lov_heapandindex(rel, comptypeOid,
&(heapOid),
&(indexOid),
heapRelfilenode,
indexRelfilenode);
START_CRIT_SECTION();
MarkBufferDirty(metabuf);
/* initialize the metapage */
PageInit(page, BufferGetPageSize(metabuf), 0);
metapage = (BMMetaPage) PageGetContents(page);
metapage->bm_magic = BITMAP_MAGIC;
metapage->bm_version = BITMAP_VERSION;
metapage->bm_lov_heapId = heapOid;
metapage->bm_lov_indexId = indexOid;
if (use_wal)
_bitmap_log_metapage(rel, page);
/* allocate the first LOV page. */
buf = _bitmap_getbuf(rel, P_NEW, BM_WRITE);
_bitmap_init_lovpage(rel, buf);
MarkBufferDirty(buf);
currLovPage = BufferGetPage(buf);
/* set the first item to support NULL value */
lovItem = _bitmap_formitem(0);
newOffset = OffsetNumberNext(PageGetMaxOffsetNumber(currLovPage));
/*
* XXX: perhaps this could be a special page, with more efficient storage
* after all, we have fixed size data
*/
o = PageAddItem(currLovPage, (Item)lovItem, sizeof(BMLOVItemData),
newOffset, LP_USED);
if (o == InvalidOffsetNumber)
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("failed to add LOV item to \"%s\"",
RelationGetRelationName(rel))));
metapage->bm_lov_lastpage = BufferGetBlockNumber(buf);
if(use_wal)
_bitmap_log_lovitem(rel, buf, newOffset, lovItem, metabuf, true);
END_CRIT_SECTION();
_bitmap_wrtbuf(buf);
_bitmap_wrtbuf(metabuf);
MIRROREDLOCK_BUFMGR_UNLOCK;
// -------- MirroredLock ----------
pfree(lovItem);
}
/*
* Backend checks if acknowledgement that its operation is completed
* is received from mirror.
* If acknowledgement is received (state == FileRepAckStateCompleted) then
* a) entry is removed from hash
* b) TRUE is returned
*/
bool
FileRepAckPrimary_IsOperationCompleted(
FileRepIdentifier_u fileRepIdentifier,
FileRepRelationType_e fileRepRelationType)
{
FileRepAckHashEntry_s *entry = NULL;
bool isCompleted = FALSE;
bool isRemoved;
FileName fileName = NULL;
int retry = 0;
bool retval = FALSE;
bool wait = FALSE;
fileName = FileRep_GetFileName(fileRepIdentifier, fileRepRelationType);
while ((isCompleted == FALSE) && FileRep_IsRetry(retry))
{
LWLockAcquire(FileRepAckHashShmemLock, LW_EXCLUSIVE);
entry = FileRepAckPrimary_LookupHashEntry(fileName);
if (entry == NULL)
{
LWLockRelease(FileRepAckHashShmemLock);
break;
}
if (! FileRep_IsIpcSleep(entry->fileRepOperation))
{
if (wait == TRUE)
{
wait = FALSE;
}
}
switch (entry->fileRepAckState)
{
case FileRepAckStateWaiting:
/* No Operation */
break;
case FileRepAckStateCompleted:
retval = TRUE;
xLogEof = entry->xLogEof;
mirrorStatus = entry->mirrorStatus;
/* no BREAK */
case FileRepAckStateMirrorInFault:
isCompleted = TRUE;
isRemoved = FileRepAckPrimary_RemoveHashEntry(fileName);
Assert(isRemoved == TRUE);
break;
default:
break;
}
if (isCompleted == false)
{
if (! FileRep_IsIpcSleep(entry->fileRepOperation))
{
fileRepIpcArray[fileRepAckHashShmem->ipcArrayIndex]->refCountSemP++;
wait = TRUE;
}
}
LWLockRelease(FileRepAckHashShmemLock);
if (isCompleted == false)
{
if (FileRepSubProcess_IsStateTransitionRequested())
{
break;
}
if (FileRep_IsIpcSleep(entry->fileRepOperation))
{
FileRep_Sleep1ms(retry);
if (retry == (3 * file_rep_retry / 4))
ereport(WARNING,
(errmsg("threshold '75' percent of 'gp_segment_connect_timeout=%d' is reached, "
"mirror may not be able to keep up with primary, "
"primary may transition to change tracking",
gp_segment_connect_timeout),
errhint("increase guc 'gp_segment_connect_timeout' by 'gpconfig' and 'gpstop -u' "),
errSendAlert(true)));
FileRep_IncrementRetry(retry);
}
else
{
FileRep_IpcWait(fileRepIpcArray[fileRepAckHashShmem->ipcArrayIndex]->semP, &fileRepIpcArray[fileRepAckHashShmem->ipcArrayIndex]->refCountSemP, FileRepAckHashShmemLock);
}
/*
* if the message was from the main filerep process then it is a
* graceful shutdown message to the mirror. We don't want to stall
* shutdown if the mirror is unavailable so we wait a smaller amount
* of time
*/
if ( entry->fileRepOperation == FileRepOperationShutdown &&
retry == 50)
{
FileRepAckPrimary_RemoveHashEntry(fileName);
break;
}
}
}
if (retval == FALSE)
{
mirrorStatus = FileRepStatusMirrorLossOccurred;
if (! primaryMirrorIsIOSuspended())
{
ereport(WARNING,
(errmsg("mirror failure, "
"could not complete mirrored request identifier '%s' ack state '%s', "
"failover requested",
(fileName == NULL) ? "<null>" : fileName,
(entry == NULL) ? "<entry not found>" : FileRepAckStateToString[entry->fileRepAckState]),
errhint("run gprecoverseg to re-establish mirror connectivity"),
FileRep_errdetail_ShmemAck(),
FileRep_errcontext()));
}
}
if (fileName)
{
pfree(fileName);
fileName = NULL;
}
return retval;
}
/* --------------------------------
* 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 = IsAutoVacuumProcess();
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();
/*
* 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;
}
/*
* FileRepAckPrimary_RunConsumer()
*/
static int
FileRepAckPrimary_RunConsumer(void)
{
FileRepShmemMessageDescr_s *fileRepShmemMessageDescr = NULL;
FileRepMessageHeader_s *fileRepMessageHeader = NULL;
pg_crc32 *fileRepMessageHeaderCrc;
pg_crc32 messageHeaderCrcLocal = 0;
int status = STATUS_OK;
bool movePositionConsume = FALSE;
FileRepShmem_s *fileRepAckShmem = NULL;
FileRep_InsertConfigLogEntry("run consumer");
fileRepAckShmem = fileRepAckShmemArray[FILEREP_ACKSHMEM_MESSAGE_SLOT_PRIMARY_ACK];
while (1) {
LWLockAcquire(FileRepAckShmemLock, LW_EXCLUSIVE);
if (movePositionConsume) {
fileRepAckShmem->positionConsume =
fileRepAckShmem->positionConsume +
fileRepShmemMessageDescr->messageLength +
sizeof(FileRepShmemMessageDescr_s);
if (fileRepAckShmem->positionConsume == fileRepAckShmem->positionWraparound &&
fileRepAckShmem->positionInsert != fileRepAckShmem->positionWraparound) {
fileRepAckShmem->positionConsume = fileRepAckShmem->positionBegin;
fileRepAckShmem->positionWraparound = fileRepAckShmem->positionEnd;
}
FileRep_IpcSignal(fileRepIpcArray[fileRepAckShmem->ipcArrayIndex]->semP,
&fileRepIpcArray[fileRepAckShmem->ipcArrayIndex]->refCountSemP);
}
fileRepShmemMessageDescr =
(FileRepShmemMessageDescr_s*) fileRepAckShmem->positionConsume;
while ((fileRepAckShmem->positionConsume == fileRepAckShmem->positionInsert) ||
((fileRepAckShmem->positionConsume != fileRepAckShmem->positionInsert) &&
(fileRepShmemMessageDescr->messageState != FileRepShmemMessageStateReady))) {
fileRepIpcArray[fileRepAckShmem->ipcArrayIndex]->refCountSemC++;
LWLockRelease(FileRepAckShmemLock);
FileRepSubProcess_ProcessSignals();
if (FileRepSubProcess_GetState() != FileRepStateReady &&
FileRepSubProcess_GetState() != FileRepStateInitialization) {
LWLockAcquire(FileRepAckShmemLock, LW_EXCLUSIVE);
break;
}
FileRep_IpcWait(fileRepIpcArray[fileRepAckShmem->ipcArrayIndex]->semC, &fileRepIpcArray[fileRepAckShmem->ipcArrayIndex]->refCountSemC, FileRepAckShmemLock);
LWLockAcquire(FileRepAckShmemLock, LW_EXCLUSIVE);
if (fileRepAckShmem->positionConsume == fileRepAckShmem->positionWraparound &&
fileRepAckShmem->positionInsert != fileRepAckShmem->positionWraparound) {
fileRepAckShmem->positionConsume = fileRepAckShmem->positionBegin;
fileRepAckShmem->positionWraparound = fileRepAckShmem->positionEnd;
}
/* Re-assign to find if messageState is changed */
fileRepShmemMessageDescr =
(FileRepShmemMessageDescr_s*) fileRepAckShmem->positionConsume;
} // internal while
fileRepAckShmem->consumeCount++;
LWLockRelease(FileRepAckShmemLock);
FileRepSubProcess_ProcessSignals();
if (FileRepSubProcess_GetState() != FileRepStateReady &&
FileRepSubProcess_GetState() != FileRepStateInitialization) {
break;
}
SIMPLE_FAULT_INJECTOR(FileRepConsumer);
/* Calculate and compare FileRepMessageHeader_s Crc */
fileRepMessageHeader = (FileRepMessageHeader_s*) (fileRepAckShmem->positionConsume +
sizeof(FileRepShmemMessageDescr_s));
FileRep_CalculateCrc((char *) fileRepMessageHeader,
sizeof(FileRepMessageHeader_s),
&messageHeaderCrcLocal);
fileRepMessageHeaderCrc =
(pg_crc32 *) (fileRepAckShmem->positionConsume +
sizeof(FileRepMessageHeader_s) +
sizeof(FileRepShmemMessageDescr_s));
if (*fileRepMessageHeaderCrc != messageHeaderCrcLocal)
{
status = STATUS_ERROR;
ereport(WARNING,
(errmsg("mirror failure, "
"could not match ack message header checksum between primary '%u' and mirror '%u', "
"failover requested",
*fileRepMessageHeaderCrc,
messageHeaderCrcLocal),
errhint("run gprecoverseg to re-establish mirror connectivity"),
FileRep_errdetail(fileRepMessageHeader->fileRepIdentifier,
fileRepMessageHeader->fileRepRelationType,
fileRepMessageHeader->fileRepOperation,
fileRepMessageHeader->messageCount),
FileRep_errdetail_ShmemAck(),
FileRep_errcontext()));
break;
}
/* Write operation is never acknowledged.
* That means message should never have body.
* CRC of body should be always 0.
*/
Assert(fileRepMessageHeader->fileRepOperation != FileRepOperationWrite);
Assert(fileRepMessageHeader->fileRepMessageBodyCrc == 0);
switch (fileRepMessageHeader->fileRepOperation)
{
case FileRepOperationReconcileXLogEof:
xLogEof = fileRepMessageHeader->fileRepOperationDescription.reconcile.xLogEof;
if (Debug_filerep_print)
ereport(LOG,
(errmsg("ack reconcile xlogid '%d' xrecoff '%d' ",
xLogEof.xlogid,
xLogEof.xrecoff)));
break;
case FileRepOperationValidation:
mirrorStatus = fileRepMessageHeader->fileRepOperationDescription.validation.mirrorStatus;
if (Debug_filerep_print)
ereport(LOG,
(errmsg("ack validation status '%s' ",
FileRepStatusToString[mirrorStatus])));
break;
case FileRepOperationCreate:
mirrorStatus = fileRepMessageHeader->fileRepOperationDescription.create.mirrorStatus;
if (Debug_filerep_print)
ereport(LOG,
(errmsg("ack create status '%s' ",
FileRepStatusToString[mirrorStatus])));
break;
case FileRepOperationStartSlruChecksum:
mirrorStatus =
fileRepMessageHeader->fileRepOperationDescription.startChecksum.mirrorStatus;
if (Debug_filerep_print)
{
ereport(LOG,
(errmsg("ack start SLRU checksum: status = '%s', directory = '%s' ",
FileRepStatusToString[mirrorStatus],
fileRepMessageHeader->fileRepIdentifier.fileRepFlatFileIdentifier.directorySimpleName)));
}
break;
case FileRepOperationVerifySlruDirectoryChecksum:
mirrorStatus =
fileRepMessageHeader->fileRepOperationDescription.verifyDirectoryChecksum.mirrorStatus;
if (Debug_filerep_print)
{
ereport(LOG,
(errmsg("ack verify SLRU directory checksum: status = '%s', directory = '%s' ",
FileRepStatusToString[mirrorStatus],
fileRepMessageHeader->fileRepIdentifier.fileRepFlatFileIdentifier.directorySimpleName)));
}
break;
default:
break;
}
if (fileRepMessageHeader->fileRepAckState != FileRepAckStateCompleted) {
status = STATUS_ERROR;
ereport(WARNING,
(errmsg("mirror failure, "
"could not complete operation on mirror ack state '%s', "
"failover requested",
FileRepAckStateToString[fileRepMessageHeader->fileRepAckState]),
errhint("run gprecoverseg to re-establish mirror connectivity"),
errSendAlert(true),
FileRep_errdetail(fileRepMessageHeader->fileRepIdentifier,
fileRepMessageHeader->fileRepRelationType,
fileRepMessageHeader->fileRepOperation,
fileRepMessageHeader->messageCount),
FileRep_errdetail_Shmem(),
FileRep_errdetail_ShmemAck(),
FileRep_errcontext()));
/*
* FAULT has to be set before entry is updated in ack hash table
* in order to suspend backend process.
*/
FileRep_SetSegmentState(SegmentStateFault, FaultTypeMirror);
FileRepSubProcess_ProcessSignals();
}
if (FileRepAckPrimary_UpdateHashEntry(
fileRepMessageHeader->fileRepIdentifier,
fileRepMessageHeader->fileRepRelationType,
fileRepMessageHeader->fileRepAckState) != STATUS_OK) {
status = STATUS_ERROR;
ereport(WARNING,
(errmsg("mirror failure, "
"could not update ack state '%s' in ack hash table, "
"failover requested",
FileRepAckStateToString[fileRepMessageHeader->fileRepAckState]),
errhint("run gprecoverseg to re-establish mirror connectivity"),
errSendAlert(true),
FileRep_errdetail(fileRepMessageHeader->fileRepIdentifier,
fileRepMessageHeader->fileRepRelationType,
fileRepMessageHeader->fileRepOperation,
fileRepMessageHeader->messageCount),
FileRep_errdetail_Shmem(),
FileRep_errdetail_ShmemAck(),
FileRep_errcontext()));
}
FileRep_InsertLogEntry(
"P_RunConsumer",
fileRepMessageHeader->fileRepIdentifier,
fileRepMessageHeader->fileRepRelationType,
fileRepMessageHeader->fileRepOperation,
messageHeaderCrcLocal,
fileRepMessageHeader->fileRepMessageBodyCrc,
fileRepMessageHeader->fileRepAckState,
FILEREP_UNDEFINED,
fileRepMessageHeader->messageCount);
if (status != STATUS_OK) {
break;
}
movePositionConsume = TRUE;
} // while(1)
return status;
}
void
disableQDMirroring(char *detail, char *errorMessage, QDMIRRORDisabledReason disabledReason)
{
struct timeval lastLogTimeVal;
Assert(ftsQDMirrorInfo != NULL);
LWLockAcquire(ftsQDMirrorLock, LW_EXCLUSIVE);
/*
* Don't overwrite.
*/
if (ftsQDMirrorInfo->state != QDMIRROR_STATE_DISABLED)
{
ftsQDMirrorInfo->state = QDMIRROR_STATE_DISABLED;
ftsQDMirrorInfo->disabledReason = disabledReason;
switch (disabledReason)
{
case QDMIRROR_DISABLEDREASON_TOOFARBEHIND:
case QDMIRROR_DISABLEDREASON_CONNECTIONERROR:
case QDMIRROR_DISABLEDREASON_WALSENDSERVERERROR:
case QDMIRROR_DISABLEDREASON_ADMINISTRATORDISABLED:
case QDMIRROR_DISABLEDREASON_UNEXPECTEDERROR:
if (disabledReason == QDMIRROR_DISABLEDREASON_ADMINISTRATORDISABLED)
ereport(NOTICE,
(errcode(ERRCODE_GP_INTERCONNECTION_ERROR),
errmsg("Master mirroring not synchronized because the standby master was not started by the administrator")));
else
ereport(WARNING,
(errcode(ERRCODE_GP_INTERCONNECTION_ERROR),
errSendAlert(true),
errmsg("Master mirroring synchronization lost"),
errdetail("%s\nThe Greenplum Database is no longer highly available.", detail)));
gettimeofday(&lastLogTimeVal, NULL);
ftsQDMirrorInfo->QDMirroringNotSynchronizedWarningGiven = true;
if (ftsQDMirrorInfo->valid)
{
/*
* Only do the update if we need to.
*/
ftsQDMirrorInfo->updateMask |= QDMIRROR_UPDATEMASK_VALIDFLAG;
}
ftsQDMirrorInfo->updateMask |= QDMIRROR_UPDATEMASK_MASTERMIRRORING;
ftsQDMirrorInfo->lastLogTimeVal = lastLogTimeVal;
/*
* Not specified or too long error message string will end up NULL
* in gp_master_mirroring table.
*/
if (errorMessage == NULL ||
strlen(errorMessage) + 1 >= sizeof(ftsQDMirrorInfo->errorMessage))
strcpy(ftsQDMirrorInfo->errorMessage, "");
else
strcpy(ftsQDMirrorInfo->errorMessage, errorMessage);
break;
case QDMIRROR_DISABLEDREASON_SHUTDOWN:
break;
default:
elog(ERROR, "Unknown disabled reason %d", (int)disabledReason);
}
}
LWLockRelease(ftsQDMirrorLock);
}
/*
* _hash_checkpage -- sanity checks on the format of all hash pages
*/
void
_hash_checkpage(Relation rel, Buffer buf, int flags)
{
Page page = BufferGetPage(buf);
/*
* ReadBuffer verifies that every newly-read page passes
* PageHeaderIsValid, which means it either contains a reasonably sane
* page header or is all-zero. We have to defend against the all-zero
* case, however.
*/
if (PageIsNew(page))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" contains unexpected zero page at block %u",
RelationGetRelationName(rel),
BufferGetBlockNumber(buf)),
errhint("Please REINDEX it."),
errSendAlert(true)));
/*
* Additionally check that the special area looks sane.
*/
if (((PageHeader) (page))->pd_special !=
(BLCKSZ - MAXALIGN(sizeof(HashPageOpaqueData))))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" contains corrupted page at block %u",
RelationGetRelationName(rel),
BufferGetBlockNumber(buf)),
errhint("Please REINDEX it."),
errSendAlert(true)));
if (flags)
{
HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
if ((opaque->hasho_flag & flags) == 0)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" contains corrupted page at block %u",
RelationGetRelationName(rel),
BufferGetBlockNumber(buf)),
errhint("Please REINDEX it."),
errSendAlert(true)));
}
/*
* When checking the metapage, also verify magic number and version.
*/
if (flags == LH_META_PAGE)
{
HashMetaPage metap = (HashMetaPage) page;
if (metap->hashm_magic != HASH_MAGIC)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" is not a hash index",
RelationGetRelationName(rel))));
if (metap->hashm_version != HASH_VERSION)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" has wrong hash version",
RelationGetRelationName(rel)),
errhint("Please REINDEX it."),
errSendAlert(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;
}
