Support separation of catalog and compute

configure

@@ -761,6 +761,7 @@ enable_gpcloud
 enable_mapreduce
 enable_orca
 enable_catalog_ext
+enable_serverless
 autodepend
 PKG_CONFIG_LIBDIR
 PKG_CONFIG_PATH
@@ -894,6 +895,7 @@ enable_depend
 enable_cassert
 enable_orca
 enable_catalog_ext
+enable_serverless
 enable_mapreduce
 enable_gpcloud
 enable_external_fts
@@ -1604,6 +1606,7 @@ Optional Features:
   --enable-cassert        enable assertion checks (for debugging)
   --disable-orca          disable ORCA optimizer
   --enable-catalog-ext    enable CloudberryDB catalog extension
+  --enable-serverless     use CloudberryDB serverless architecture
   --enable-mapreduce      enable CloudberryDB Mapreduce support
   --enable-gpcloud        enable gpcloud support
   --enable-external-fts   enable external fts support
@@ -8340,6 +8343,37 @@ fi
 
 { $as_echo "$as_me:${as_lineno-$LINENO}: result: checking whether to build with CloudberryDB catalog extension... $enable_catalog_ext" >&5
 $as_echo "checking whether to build with CloudberryDB catalog extension ... $enable_catalog_ext" >&6; }
+#
+# Enable serverless architecture
+#
+
+
+# Check whether --enable-serverless was given.
+if test "${enable_serverless+set}" = set; then :
+  enableval=$enable_serverless;
+  case $enableval in
+    yes)
+
+$as_echo "#define SERVERLESS 1" >>confdefs.h
+
+      ;;
+    no)
+      :
+      ;;
+    *)
+      as_fn_error $? "no argument expected for --enable-serverless option" "$LINENO" 5
+      ;;
+  esac
+
+else
+  enable_serverless=no
+
+fi
+
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: checking whether to use serverless architecture of Cloudberry ... $enable_serverless" >&5
+$as_echo "checking whether to use serverless architecture of Cloudberry ... $enable_serverless" >&6; }
+
 
 #
 # --enable-mapreduce enables GPMapreduce support

configure.ac

@@ -846,6 +846,16 @@ PGAC_ARG_BOOL(enable, catalog-ext, no, [enable Cloudberry catalog extension],
 AC_MSG_RESULT([checking whether to build with catalog extension... $enable_catalog_ext])
 AC_SUBST(enable_catalog_ext)
 
+#
+#
+# --enable-serverless uses serverless architecture of Cloudberry
+#
+PGAC_ARG_BOOL(enable, serverless, no, [use serverless architecture of Cloudberry],
+              [AC_DEFINE([SERVERLESS], 1,
+                                    [Define to 1 to use serverless architecture of Cloudberry. (--enable-serverless)])])
+AC_MSG_RESULT([checking whether to use serverless architecture of Cloudberry... $enable_serverless])
+AC_SUBST(enable_serverless)
+
 #
 # --enable-mapreduce enables GPMapreduce support
 #

contrib/interconnect/udp/ic_udpifc.c

@@ -1599,6 +1599,14 @@ initConnHashTable(ConnHashTable *ht, MemoryContext cxt)
 
 	ht->cxt = cxt;
 	ht->size = Gp_role == GP_ROLE_DISPATCH ? (getgpsegmentCount() * 2) : ic_htab_size;
+
+	/*
+	 * In serverless architecture, the cluster may have only one QD, skip Initialization.
+	 * Initialization will be done later.
+	 */
+	if (enable_serverless && Gp_role == GP_ROLE_DISPATCH && ht->size == 0)
+		return true;
+
 	Assert(ht->size > 0);
 
 	if (ht->cxt)
@@ -1635,6 +1643,18 @@ connAddHash(ConnHashTable *ht, MotionConn *mConn)
 	MemoryContext old = NULL;
 	MotionConnUDP *conn = NULL;
 
+	/*
+	 * Initialize connection hash table if needed.
+	 */
+	if (enable_serverless && Gp_role == GP_ROLE_DISPATCH && ht->size == 0)
+	{
+		old = MemoryContextSwitchTo(ht->cxt);
+		initConnHashTable(ht, ht->cxt);
+		MemoryContextSwitchTo(old);
+
+		Assert(ht->size > 0);
+	}
+
 	conn = CONTAINER_OF(mConn, MotionConnUDP, mConn);
 
 	hashcode = CONN_HASH_VALUE(&conn->conn_info) % ht->size;
@@ -1805,10 +1825,13 @@ destroyConnHashTable(ConnHashTable *ht)
 		}
 	}
 
-	if (ht->cxt)
-		pfree(ht->table);
-	else
-		free(ht->table);
+	if (ht->size > 0)
+	{
+		if (ht->cxt)
+			pfree(ht->table);
+		else
+			free(ht->table);
+	}
 
 	ht->table = NULL;
 	ht->size = 0;

src/Makefile.global.in

@@ -223,6 +223,7 @@ enable_strong_random	= @enable_strong_random@
 enable_largefile	= @enable_largefile@
 enable_orca		= @enable_orca@
 enable_catalog_ext  = @enable_catalog_ext@
+enable_serverless = @enable_serverless@
 enable_gpfdist		= @enable_gpfdist@
 enable_pxf			= @enable_pxf@
 enable_debug_extensions = @enable_debug_extensions@

src/backend/Makefile

@@ -32,6 +32,9 @@ ifeq ($(enable_catalog_ext),yes)
 SUBDIRS += catalog-extension
 LDFLAGS += -lprotobuf -lstdc++
 endif
+ifeq ($(enable_serverless),yes)
+LDFLAGS += -lprotobuf -lstdc++ -ljansson
+endif
 
 include $(srcdir)/common.mk
 

src/backend/access/transam/clog.c

@@ -38,6 +38,7 @@
 #include "access/xlog.h"
 #include "access/xloginsert.h"
 #include "access/xlogutils.h"
+#include "cdb/cdbvars.h"
 #include "miscadmin.h"
 #include "pg_trace.h"
 #include "pgstat.h"
@@ -167,6 +168,12 @@ TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
 	int			pageno = TransactionIdToPage(xid);	/* get page of parent */
 	int			i;
 
+	/*
+	 * Only master can set transaction status
+	 */
+	if (enable_serverless && (Gp_role != GP_ROLE_DISPATCH && GpIdentity.segindex != MASTER_CONTENT_ID))
+		return;
+
 	Assert(status == TRANSACTION_STATUS_COMMITTED ||
 		   status == TRANSACTION_STATUS_ABORTED);
 

src/backend/access/transam/slru.c

@@ -133,16 +133,16 @@ typedef enum
 static SlruErrorCause slru_errcause;
 static int	slru_errno;
 
+/*
+ * Hooks for plugins to get control in SlruPhysicalReadPage/SlruPhysicalWritePage/SimpleLruReadPage
+ */
+SlruPhysicalReadPage_hook_type SlruPhysicalReadPage_hook = NULL;
+SlruPhysicalWritePage_hook_type SlruPhysicalWritePage_hook = NULL;
+SimpleLruReadPage_hook_type SimpleLruReadPage_hook = NULL;
 
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
 static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
 static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
 								  SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int	SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
 static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
 									  int segpage, void *data);
 static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
@@ -319,7 +319,7 @@ SimpleLruZeroPage(SlruCtl ctl, int pageno)
  *
  * This assumes that InvalidXLogRecPtr is bitwise-all-0.
  */
-static void
+void
 SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
 {
 	SlruShared	shared = ctl->shared;
@@ -336,7 +336,7 @@ SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
  *
  * Control lock must be held at entry, and will be held at exit.
  */
-static void
+void
 SimpleLruWaitIO(SlruCtl ctl, int slotno)
 {
 	SlruShared	shared = ctl->shared;
@@ -396,6 +396,9 @@ SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
 {
 	SlruShared	shared = ctl->shared;
 
+	if (SimpleLruReadPage_hook)
+		return (*SimpleLruReadPage_hook) (ctl, pageno, write_ok, xid);
+
 	/* Outer loop handles restart if we must wait for someone else's I/O */
 	for (;;)
 	{
@@ -496,6 +499,12 @@ SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
 	SlruShared	shared = ctl->shared;
 	int			slotno;
 
+	if (SimpleLruReadPage_hook)
+	{
+		LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
+		return (*SimpleLruReadPage_hook) (ctl, pageno, true, xid);
+	}
+
 	/* Try to find the page while holding only shared lock */
 	LWLockAcquire(shared->ControlLock, LW_SHARED);
 
@@ -679,7 +688,7 @@ SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
  * For now, assume it's not worth keeping a file pointer open across
  * read/write operations.  We could cache one virtual file pointer ...
  */
-static bool
+bool
 SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
 {
 	SlruShared	shared = ctl->shared;
@@ -688,6 +697,13 @@ SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
 	off_t		offset = rpageno * BLCKSZ;
 	char		path[MAXPGPATH];
 	int			fd;
+	bool		result;
+
+	if (SlruPhysicalReadPage_hook &&
+		SlruPhysicalReadPage_hook(ctl, pageno, slotno, &result))
+	{
+		return result;
+	}
 
 	SlruFileName(ctl, path, segno);
 
@@ -760,6 +776,7 @@ SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
 	off_t		offset = rpageno * BLCKSZ;
 	char		path[MAXPGPATH];
 	int			fd = -1;
+	bool		result;
 
 	/* update the stats counter of written pages */
 	pgstat_count_slru_page_written(shared->slru_stats_idx);
@@ -806,6 +823,12 @@ SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
 		}
 	}
 
+	if (SlruPhysicalWritePage_hook && 
+		SlruPhysicalWritePage_hook(ctl, pageno, slotno, &result))
+	{
+		return result;
+	}
+
 	/*
 	 * During a WriteAll, we may already have the desired file open.
 	 */
@@ -926,7 +949,7 @@ SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
  * Issue the error message after failure of SlruPhysicalReadPage or
  * SlruPhysicalWritePage.  Call this after cleaning up shared-memory state.
  */
-static void
+void
 SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
 {
 	int			segno = pageno / SLRU_PAGES_PER_SEGMENT;
@@ -1011,7 +1034,7 @@ SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
  *
  * Control lock must be held at entry, and will be held at exit.
  */
-static int
+int
 SlruSelectLRUPage(SlruCtl ctl, int pageno)
 {
 	SlruShared	shared = ctl->shared;

src/backend/access/transam/subtrans.c

@@ -379,3 +379,12 @@ SubTransPagePrecedes(int page1, int page2)
 	return (TransactionIdPrecedes(xid1, xid2) &&
 			TransactionIdPrecedes(xid1, xid2 + SUBTRANS_XACTS_PER_PAGE - 1));
 }
+
+/*
+ * Get SUBTRANS control data
+ */
+SlruCtl
+SUBTRANS_Ctl(void)
+{
+    return SubTransCtl;
+}

src/backend/access/transam/varsup.c

@@ -43,6 +43,11 @@ int xid_warn_limit;
 
 NewSegRelfilenode_assign_hook_type NewSegRelfilenode_assign_hook = NULL;
 
+/*
+ * Hook for plugins to get control in GetNewTransactionId.
+ */
+GetNewTransactionId_hook_type GetNewTransactionId_hook = NULL;
+
 /*
  * Allocate the next FullTransactionId for a new transaction or
  * subtransaction.
@@ -61,6 +66,9 @@ GetNewTransactionId(bool isSubXact)
 	FullTransactionId full_xid;
 	TransactionId xid;
 
+	if (GetNewTransactionId_hook)
+		return (*GetNewTransactionId_hook) (isSubXact);
+
 	/*
 	 * Workers synchronize transaction state at the beginning of each parallel
 	 * operation, so we can't account for new XIDs after that point.