Add guc hot_dr for hot standby

.github/workflows/build-cloudberry.yml

@@ -310,6 +310,9 @@ jobs:
               {"test":"ic-isolation2",
                "make_configs":["src/test/isolation2:installcheck-isolation2"]
               },
+              {"test":"ic-isolation2-hot-standby",
+               "make_configs":["src/test/isolation2:installcheck-hot-standby"]
+              },
               {"test":"ic-isolation2-crash",
                "make_configs":["src/test/isolation2:installcheck-isolation2-crash"],
                "enable_core_check":false

src/backend/access/transam/xlog.c

@@ -114,6 +114,7 @@ int			XLogArchiveTimeout = 0;
 int			XLogArchiveMode = ARCHIVE_MODE_OFF;
 char	   *XLogArchiveCommand = NULL;
 bool		EnableHotStandby = false;
+bool		EnableHotDR = false;
 bool		fullPageWrites = true;
 bool		wal_log_hints = false;
 bool		wal_compression = false;
@@ -7967,6 +7968,12 @@ StartupXLOG(void)
 				if (gp_pause_on_restore_point_replay)
 					pauseRecoveryOnRestorePoint(xlogreader);
 
+				/* Exit the recovery loop if a promotion is triggered in pauseRecoveryOnRestorePoint() */
+				if (reachedContinuousRecoveryTarget && recoveryTargetAction == RECOVERY_TARGET_ACTION_PROMOTE){
+					reachedRecoveryTarget = true;
+					break;
+				}
+
 				/* Exit loop if we reached inclusive recovery target */
 				if (recoveryStopsAfter(xlogreader))
 				{
@@ -10757,6 +10764,9 @@ XLogRestorePoint(const char *rpName)
 	xlrec.rp_time = GetCurrentTimestamp();
 	strlcpy(xlrec.rp_name, rpName, MAXFNAMELEN);
 
+	/* LogHotStandby for the restore here */
+	LogStandbySnapshot();
+
 	XLogBeginInsert();
 	XLogRegisterData((char *) &xlrec, sizeof(xl_restore_point));
 

src/backend/cdb/cdbutil.c

@@ -92,6 +92,7 @@ static int	CdbComponentDatabaseInfoCompare(const void *p1, const void *p2);
 
 static GpSegConfigEntry * readGpSegConfigFromCatalog(int *total_dbs);
 static GpSegConfigEntry * readGpSegConfigFromFTSFiles(int *total_dbs);
+static GpSegConfigEntry * readGpSegConfigFromFiles(int *total_dbs);
 
 static void getAddressesForDBid(GpSegConfigEntry *c, int elevel);
 static HTAB *hostPrimaryCountHashTableInit(void);
@@ -131,6 +132,15 @@ typedef struct HostPrimaryCountEntry
  */
 static GpSegConfigEntry *
 readGpSegConfigFromFTSFiles(int *total_dbs)
+{
+	Assert(!IsTransactionState() && !IS_HOT_DR_CLUSTER());
+	/* notify and wait FTS to finish a probe and update the dump file */
+	FtsNotifyProber();
+	return readGpSegConfigFromFiles(total_dbs);
+}
+
+static GpSegConfigEntry *
+readGpSegConfigFromFiles(int *total_dbs)
 {
 	FILE	*fd;
 	int		idx = 0;
@@ -142,11 +152,6 @@ readGpSegConfigFromFTSFiles(int *total_dbs)
 	char	address[MAXHOSTNAMELEN];
 	char	buf[MAXHOSTNAMELEN * 2 + 32];
 
-	Assert(!IsTransactionState());
-
-	/* notify and wait FTS to finish a probe and update the dump file */
-	FtsNotifyProber();	
-
 	fd = AllocateFile(GPSEGCONFIGDUMPFILE, "r");
 
 	if (!fd)
@@ -188,6 +193,18 @@ readGpSegConfigFromFTSFiles(int *total_dbs)
 	return configs;
 }
 
+bool
+checkGpSegConfigFtsFiles()
+{
+	FILE *fd = AllocateFile(GPSEGCONFIGDUMPFILE, "r");
+
+	if (!fd)
+		return false;
+
+	FreeFile(fd);
+	return true;
+}
+
 /*
  * writeGpSegConfigToFTSFiles() dump gp_segment_configuration to the file
  * GPSEGCONFIGDUMPFILE, in $PGDATA, only FTS process can use this function.
@@ -372,10 +389,17 @@ getCdbComponentInfo(void)
 
 	HTAB	   *hostPrimaryCountHash = hostPrimaryCountHashTableInit();
 
-	if (IsTransactionState())
-		configs = readGpSegConfigFromCatalog(&total_dbs);
+	if (EnableHotDR)
+	{
+		configs = readGpSegConfigFromFiles(&total_dbs);
+	}
 	else
-		configs = readGpSegConfigFromFTSFiles(&total_dbs);
+	{
+		if (IsTransactionState())
+			configs = readGpSegConfigFromCatalog(&total_dbs);
+		else
+			configs = readGpSegConfigFromFTSFiles(&total_dbs);
+	}
 
 	component_databases = palloc0(sizeof(CdbComponentDatabases));
 

src/backend/utils/misc/guc_gp.c

@@ -86,6 +86,7 @@ static bool check_optimizer(bool *newval, void **extra, GucSource source);
 static bool check_verify_gpfdists_cert(bool *newval, void **extra, GucSource source);
 static bool check_dispatch_log_stats(bool *newval, void **extra, GucSource source);
 static bool check_gp_workfile_compression(bool *newval, void **extra, GucSource source);
+static bool check_hot_dr(bool *newval, void **extra, GucSource source);
 
 /* Helper function for guc setter */
 bool gpvars_check_gp_resqueue_priority_default_value(char **newval,
@@ -3331,6 +3332,16 @@ struct config_bool ConfigureNamesBool_gp[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"hot_dr", PGC_POSTMASTER, REPLICATION_STANDBY,
+			gettext_noop("DR Cluster as well as allows connteions and queries"),
+			NULL
+		},
+		&EnableHotDR,
+		false,
+		check_hot_dr, NULL, NULL
+	},
+
 	{
 		{"gp_enable_runtime_filter_pushdown", PGC_USERSET, DEVELOPER_OPTIONS,
 			gettext_noop("Try to push the hash table of hash join to the seqscan or AM as bloom filter."),
@@ -5455,6 +5466,22 @@ check_verify_gpfdists_cert(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_hot_dr(bool *newval, void **extra, GucSource source)
+{
+	if (*newval && !EnableHotStandby)
+		ereport(ERROR,
+				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+				 errmsg("cannot enable \"hot_dr\" when \"hot_standby\" is false")));
+
+	if (*newval && IS_QUERY_DISPATCHER() && !checkGpSegConfigFtsFiles())
+		ereport(ERROR,
+				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+				 errmsg("cannot enable \"hot_dr\" since DR cluster segment configuration file does not exits")));
+
+	return true;
+}
+
 static bool
 check_dispatch_log_stats(bool *newval, void **extra, GucSource source)
 {

src/include/access/xlog.h

@@ -123,6 +123,7 @@ extern int	XLogArchiveTimeout;
 extern int	wal_retrieve_retry_interval;
 extern char *XLogArchiveCommand;
 extern bool EnableHotStandby;
+extern bool EnableHotDR;
 
 extern bool fullPageWrites;
 extern bool wal_log_hints;

src/include/cdb/cdbutil.h

@@ -132,6 +132,7 @@ extern char *getDnsAddress(char *name, int port, int elevel);
 
 #ifdef USE_INTERNAL_FTS
 extern void writeGpSegConfigToFTSFiles(void);
+extern bool checkGpSegConfigFtsFiles(void);
 #else
 
 GpSegConfigEntry * readGpSegConfig(char * buff, int *total_dbs);

src/include/cdb/cdbvars.h

@@ -757,6 +757,7 @@ extern GpId GpIdentity;
 #define MAX_DBID_STRING_LENGTH  11
 
 #define UNINITIALIZED_GP_IDENTITY_VALUE (-10000)
+#define IS_HOT_DR_CLUSTER() (EnableHotDR)
 #define IS_QUERY_DISPATCHER() (GpIdentity.segindex == MASTER_CONTENT_ID)
 #define IS_HOT_STANDBY_QD() (EnableHotStandby && IS_QUERY_DISPATCHER() && RecoveryInProgress())
 

src/include/utils/unsync_guc_name.h

@@ -294,6 +294,7 @@
 		"gp_workfile_limit_per_segment",
 		"gp_workfile_max_entries",
 		"hba_file",
+		"hot_dr",
 		"hot_standby",
 		"hot_standby_feedback",
 		"huge_pages",

src/test/isolation2/expected/hot_standby/faults.out

@@ -133,14 +133,15 @@ select gp_inject_fault('out_of_recovery_in_startupxlog', 'reset', dbid) from gp_
 ERROR:  primary segments can only process MPP protocol messages from primary QD  (seg1 slice1 127.0.1.1:7006 pid=14671)
 HINT:  Exit the current session and re-connect.
 -1Sq: ... <quitting>
-
+-- start_ignore
 -- will fail due to downed mirror (previous primary)
 -1S: select * from hs_failover;
 ERROR:  failed to acquire resources on one or more segments
 DETAIL:  connection to server at "10.13.9.74", port 7003 failed: Connection refused
 	Is the server running on that host and accepting TCP/IP connections?
  (seg1 10.13.9.74:7003)
 -1Sq: ... <quitting>
+-- end_ignore
 
 -- bring the downed mirror up
 !\retcode gprecoverseg -aF;

src/test/isolation2/sql/hot_standby/faults.sql

@@ -59,10 +59,11 @@ select gp_inject_fault('out_of_recovery_in_startupxlog', 'reset', dbid) from gp_
 -- in an existing gang. That mirror is now a primary, so it will complain and the query fails.
 -1S: select * from hs_failover;
 -1Sq:
-
+-- start_ignore
 -- will fail due to downed mirror (previous primary)
 -1S: select * from hs_failover;
 -1Sq:
+-- end_ignore
 
 -- bring the downed mirror up
 !\retcode gprecoverseg -aF;