Add parallel pg_dump option.

New infrastructure is added which creates a set number of workers
(threads on Windows, forked processes on Unix). Jobs are then
handed out to these workers by the master process as needed.
pg_restore is adjusted to use this new infrastructure in place of the
old setup which created a new worker for each step on the fly. Parallel
dumps acquire a snapshot clone in order to stay consistent, if
available.

The parallel option is selected by the -j / --jobs command line
parameter of pg_dump.

Joachim Wieland, lightly editorialized by Andrew Dunstan.

doc/src/sgml/backup.sgml

@@ -310,6 +310,24 @@ pg_restore -d <replaceable class="parameter">dbname</replaceable> <replaceable c
     with one of the other two approaches.
    </para>
 
+   <formalpara>
+    <title>Use <application>pg_dump</>'s parallel dump feature.</title>
+    <para>
+     To speed up the dump of a large database, you can use
+     <application>pg_dump</application>'s parallel mode. This will dump
+     multiple tables at the same time. You can control the degree of
+     parallelism with the <command>-j</command> parameter. Parallel dumps
+     are only supported for the "directory" archive format.
+
+<programlisting>
+pg_dump -j <replaceable class="parameter">num</replaceable> -F d -f <replaceable class="parameter">out.dir</replaceable> <replaceable class="parameter">dbname</replaceable>
+</programlisting>
+
+     You can use <command>pg_restore -j</command> to restore a dump in parallel.
+     This will work for any archive of either the "custom" or the "directory"
+     archive mode, whether or not it has been created with <command>pg_dump -j</command>.
+    </para>
+   </formalpara>
   </sect2>
  </sect1>
 

doc/src/sgml/perform.sgml

@@ -1433,6 +1433,15 @@ SELECT * FROM x, y, a, b, c WHERE something AND somethingelse;
        base backup.
       </para>
      </listitem>
+     <listitem>
+      <para>
+       Experiment with the parallel dump and restore modes of both
+       <application>pg_dump</> and <application>pg_restore</> and find the
+       optimal number of concurrent jobs to use. Dumping and restoring in
+       parallel by means of the <option>-j</> option should give you a
+       significantly higher performance over the serial mode.
+      </para>
+     </listitem>
      <listitem>
       <para>
        Consider whether the whole dump should be restored as a single

doc/src/sgml/ref/pg_dump.sgml

@@ -73,10 +73,12 @@ PostgreSQL documentation
    transfer mechanism. <application>pg_dump</application> can be used to
    backup an entire database, then <application>pg_restore</application>
    can be used to examine the archive and/or select which parts of the
-   database are to be restored. The most flexible output file format is
+   database are to be restored. The most flexible output file formats are
-   the <quote>custom</quote> format (<option>-Fc</option>). It allows
+   the <quote>custom</quote> format (<option>-Fc</option>) and the
-   for selection and reordering of all archived items, and is compressed
+   <quote>directory</quote> format(<option>-Fd</option>). They allow
-   by default.
+   for selection and reordering of all archived items, support parallel
+   restoration, and are compressed by default. The <quote>directory</quote>
+   format is the only format that supports parallel dumps.
   </para>
 
   <para>
@@ -251,7 +253,8 @@ PostgreSQL documentation
            can read. A directory format archive can be manipulated with
            standard Unix tools; for example, files in an uncompressed archive
            can be compressed with the <application>gzip</application> tool.
-           This format is compressed by default.
+           This format is compressed by default and also supports parallel
+           dumps.
           </para>
          </listitem>
         </varlistentry>
@@ -285,6 +288,62 @@ PostgreSQL documentation
       </listitem>
      </varlistentry>
 
+     <varlistentry>
+      <term><option>-j <replaceable class="parameter">njobs</replaceable></></term>
+      <term><option>--jobs=<replaceable class="parameter">njobs</replaceable></></term>
+      <listitem>
+       <para>
+        Run the dump in parallel by dumping <replaceable class="parameter">njobs</replaceable>
+        tables simultaneously. This option reduces the time of the dump but it also
+        increases the load on the database server. You can only use this option with the
+        directory output format because this is the only output format where multiple processes
+        can write their data at the same time.
+       </para>
+       <para>
+        <application>pg_dump</> will open <replaceable class="parameter">njobs</replaceable>
+        + 1 connections to the database, so make sure your <xref linkend="guc-max-connections">
+        setting is high enough to accommodate all connections.
+       </para>
+       <para>
+        Requesting exclusive locks on database objects while running a parallel dump could
+        cause the dump to fail. The reason is that the <application>pg_dump</> master process
+        requests shared locks on the objects that the worker processes are going to dump later
+        in order to
+        make sure that nobody deletes them and makes them go away while the dump is running.
+        If another client then requests an exclusive lock on a table, that lock will not be
+        granted but will be queued waiting for the shared lock of the master process to be
+        released.. Consequently any other access to the table will not be granted either and
+        will queue after the exclusive lock request. This includes the worker process trying
+        to dump the table. Without any precautions this would be a classic deadlock situation.
+        To detect this conflict, the <application>pg_dump</> worker process requests another
+        shared lock using the <literal>NOWAIT</> option. If the worker process is not granted
+        this shared lock, somebody else must have requested an exclusive lock in the meantime
+        and there is no way to continue with the dump, so <application>pg_dump</> has no choice
+        but to abort the dump.
+       </para>
+       <para>
+        For a consistent backup, the database server needs to support synchronized snapshots,
+        a feature that was introduced in <productname>PostgreSQL</productname> 9.2. With this
+        feature, database clients can ensure they see the same dataset even though they use
+        different connections. <command>pg_dump -j</command> uses multiple database 
+        connections; it connects to the database once with the master process and
+        once again for each worker job. Without the sychronized snapshot feature, the
+        different worker jobs wouldn't be guaranteed to see the same data in each connection,
+        which could lead to an inconsistent backup.
+       </para>
+       <para>
+        If you want to run a parallel dump of a pre-9.2 server, you need to make sure that the
+        database content doesn't change from between the time the master connects to the
+        database until the last worker job has connected to the database. The easiest way to
+        do this is to halt any data modifying processes (DDL and DML) accessing the database
+        before starting the backup. You also need to specify the
+        <option>--no-synchronized-snapshots</option> parameter when running
+        <command>pg_dump -j</command> against a pre-9.2 <productname>PostgreSQL</productname>
+        server.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry>
       <term><option>-n <replaceable class="parameter">schema</replaceable></option></term>
       <term><option>--schema=<replaceable class="parameter">schema</replaceable></option></term>
@@ -690,6 +749,17 @@ PostgreSQL documentation
       </listitem>
      </varlistentry>
 
+     <varlistentry>
+      <term><option>--no-synchronized-snapshots</></term>
+      <listitem>
+       <para>
+        This option allows running <command>pg_dump -j</> against a pre-9.2
+        server, see the documentation of the <option>-j</option> parameter
+        for more details.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry>
       <term><option>--no-tablespaces</option></term>
       <listitem>
@@ -1082,6 +1152,15 @@ CREATE DATABASE foo WITH TEMPLATE template0;
 </screen>
   </para>
 
+  <para>
+   To dump a database into a directory-format archive in parallel with
+   5 worker jobs:
+
+<screen>
+<prompt>$</prompt> <userinput>pg_dump -Fd mydb -j 5 -f dumpdir</userinput>
+</screen>
+  </para>
+
   <para>
    To reload an archive file into a (freshly created) database named
    <literal>newdb</>:

src/bin/pg_dump/Makefile

@@ -19,7 +19,7 @@ include $(top_builddir)/src/Makefile.global
 override CPPFLAGS := -I$(libpq_srcdir) $(CPPFLAGS)
 
 OBJS=	pg_backup_archiver.o pg_backup_db.o pg_backup_custom.o \
-	pg_backup_null.o pg_backup_tar.o \
+	pg_backup_null.o pg_backup_tar.o parallel.o \
 	pg_backup_directory.o dumputils.o compress_io.o $(WIN32RES)
 
 KEYWRDOBJS = keywords.o kwlookup.o

src/bin/pg_dump/compress_io.c

@@ -54,6 +54,7 @@
 
 #include "compress_io.h"
 #include "dumputils.h"
+#include "parallel.h"
 
 /*----------------------
  * Compressor API
@@ -182,6 +183,9 @@ size_t
 WriteDataToArchive(ArchiveHandle *AH, CompressorState *cs,
 				   const void *data, size_t dLen)
 {
+	/* Are we aborting? */
+	checkAborting(AH);
+
 	switch (cs->comprAlg)
 	{
 		case COMPR_ALG_LIBZ:
@@ -351,6 +355,9 @@ ReadDataFromArchiveZlib(ArchiveHandle *AH, ReadFunc readF)
 	/* no minimal chunk size for zlib */
 	while ((cnt = readF(AH, &buf, &buflen)))
 	{
+		/* Are we aborting? */
+		checkAborting(AH);
+
 		zp->next_in = (void *) buf;
 		zp->avail_in = cnt;
 
@@ -411,6 +418,9 @@ ReadDataFromArchiveNone(ArchiveHandle *AH, ReadFunc readF)
 
 	while ((cnt = readF(AH, &buf, &buflen)))
 	{
+		/* Are we aborting? */
+		checkAborting(AH);
+
 		ahwrite(buf, 1, cnt, AH);
 	}
 

src/bin/pg_dump/dumputils.c

@@ -38,6 +38,7 @@ static struct
 }	on_exit_nicely_list[MAX_ON_EXIT_NICELY];
 
 static int	on_exit_nicely_index;
+void		(*on_exit_msg_func) (const char *modulename, const char *fmt, va_list ap) = vwrite_msg;
 
 #define supports_grant_options(version) ((version) >= 70400)
 
@@ -48,11 +49,21 @@ static bool parseAclItem(const char *item, const char *type,
 static char *copyAclUserName(PQExpBuffer output, char *input);
 static void AddAcl(PQExpBuffer aclbuf, const char *keyword,
 	   const char *subname);
+static PQExpBuffer getThreadLocalPQExpBuffer(void);
 
 #ifdef WIN32
+static void shutdown_parallel_dump_utils(int code, void *unused);
 static bool parallel_init_done = false;
 static DWORD tls_index;
 static DWORD mainThreadId;
+
+static void
+shutdown_parallel_dump_utils(int code, void *unused)
+{
+	/* Call the cleanup function only from the main thread */
+	if (mainThreadId == GetCurrentThreadId())
+		WSACleanup();
+}
 #endif
 
 void
@@ -61,23 +72,29 @@ init_parallel_dump_utils(void)
 #ifdef WIN32
 	if (!parallel_init_done)
 	{
+		WSADATA		wsaData;
+		int			err;
+
 		tls_index = TlsAlloc();
-		parallel_init_done = true;
 		mainThreadId = GetCurrentThreadId();
+		err = WSAStartup(MAKEWORD(2, 2), &wsaData);
+		if (err != 0)
+		{
+			fprintf(stderr, _("WSAStartup failed: %d\n"), err);
+			exit_nicely(1);
+		}
+		on_exit_nicely(shutdown_parallel_dump_utils, NULL);
+		parallel_init_done = true;
 	}
 #endif
 }
 
 /*
- *	Quotes input string if it's not a legitimate SQL identifier as-is.
+ * Non-reentrant but reduces memory leakage. (On Windows the memory leakage
- *
+ * will be one buffer per thread, which is at least better than one per call).
- *	Note that the returned string must be used before calling fmtId again,
- *	since we re-use the same return buffer each time.  Non-reentrant but
- *	reduces memory leakage. (On Windows the memory leakage will be one buffer
- *	per thread, which is at least better than one per call).
  */
-const char *
+static PQExpBuffer
-fmtId(const char *rawid)
+getThreadLocalPQExpBuffer(void)
 {
 	/*
 	 * The Tls code goes awry if we use a static var, so we provide for both
@@ -86,9 +103,6 @@ fmtId(const char *rawid)
 	static PQExpBuffer s_id_return = NULL;
 	PQExpBuffer id_return;
 
-	const char *cp;
-	bool		need_quotes = false;
-
 #ifdef WIN32
 	if (parallel_init_done)
 		id_return = (PQExpBuffer) TlsGetValue(tls_index);		/* 0 when not set */
@@ -118,6 +132,23 @@ fmtId(const char *rawid)
 
 	}
 
+	return id_return;
+}
+
+/*
+ *	Quotes input string if it's not a legitimate SQL identifier as-is.
+ *
+ *	Note that the returned string must be used before calling fmtId again,
+ *	since we re-use the same return buffer each time.
+ */
+const char *
+fmtId(const char *rawid)
+{
+	PQExpBuffer id_return = getThreadLocalPQExpBuffer();
+
+	const char *cp;
+	bool		need_quotes = false;
+
 	/*
 	 * These checks need to match the identifier production in scan.l. Don't
 	 * use islower() etc.
@@ -185,6 +216,35 @@ fmtId(const char *rawid)
 	return id_return->data;
 }
 
+/*
+ * fmtQualifiedId - convert a qualified name to the proper format for
+ * the source database.
+ *
+ * Like fmtId, use the result before calling again.
+ *
+ * Since we call fmtId and it also uses getThreadLocalPQExpBuffer() we cannot
+ * use it until we're finished with calling fmtId().
+ */
+const char *
+fmtQualifiedId(int remoteVersion, const char *schema, const char *id)
+{
+	PQExpBuffer id_return;
+	PQExpBuffer lcl_pqexp = createPQExpBuffer();
+
+	/* Suppress schema name if fetching from pre-7.3 DB */
+	if (remoteVersion >= 70300 && schema && *schema)
+	{
+		appendPQExpBuffer(lcl_pqexp, "%s.", fmtId(schema));
+	}
+	appendPQExpBuffer(lcl_pqexp, "%s", fmtId(id));
+
+	id_return = getThreadLocalPQExpBuffer();
+
+	appendPQExpBuffer(id_return, "%s", lcl_pqexp->data);
+	destroyPQExpBuffer(lcl_pqexp);
+
+	return id_return->data;
+}
 
 /*
  * Convert a string value to an SQL string literal and append it to
@@ -1315,7 +1375,7 @@ exit_horribly(const char *modulename, const char *fmt,...)
 	va_list		ap;
 
 	va_start(ap, fmt);
-	vwrite_msg(modulename, fmt, ap);
+	on_exit_msg_func(modulename, fmt, ap);
 	va_end(ap);
 
 	exit_nicely(1);

src/bin/pg_dump/dumputils.h

@@ -47,6 +47,8 @@ extern const char *progname;
 
 extern void init_parallel_dump_utils(void);
 extern const char *fmtId(const char *identifier);
+extern const char *fmtQualifiedId(int remoteVersion,
+			   const char *schema, const char *id);
 extern void appendStringLiteral(PQExpBuffer buf, const char *str,
 					int encoding, bool std_strings);
 extern void appendStringLiteralConn(PQExpBuffer buf, const char *str,
@@ -85,11 +87,12 @@ __attribute__((format(PG_PRINTF_ATTRIBUTE, 2, 0)));
 extern void
 exit_horribly(const char *modulename, const char *fmt,...)
 __attribute__((format(PG_PRINTF_ATTRIBUTE, 2, 3), noreturn));
+extern void (*on_exit_msg_func) (const char *modulename, const char *fmt, va_list ap)
+			__attribute__((format(PG_PRINTF_ATTRIBUTE, 2, 0)));
 extern void on_exit_nicely(on_exit_nicely_callback function, void *arg);
 extern void exit_nicely(int code) __attribute__((noreturn));
 
 extern void simple_string_list_append(SimpleStringList *list, const char *val);
 extern bool simple_string_list_member(SimpleStringList *list, const char *val);
 
-
 #endif   /* DUMPUTILS_H */

src/bin/pg_dump/parallel.h

@@ -0,0 +1,85 @@
+/*-------------------------------------------------------------------------
+ *
+ * parallel.h
+ *
+ *	Parallel support header file for the pg_dump archiver
+ *
+ * Portions Copyright (c) 1996-2011, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
+ *
+ *	The author is not responsible for loss or damages that may
+ *	result from its use.
+ *
+ * IDENTIFICATION
+ *		src/bin/pg_dump/parallel.h
+ *
+ *-------------------------------------------------------------------------
+ */
+
+#include "pg_backup_db.h"
+
+struct _archiveHandle;
+struct _tocEntry;
+
+typedef enum
+{
+	WRKR_TERMINATED = 0,
+	WRKR_IDLE,
+	WRKR_WORKING,
+	WRKR_FINISHED
+}	T_WorkerStatus;
+
+typedef enum T_Action
+{
+	ACT_DUMP,
+	ACT_RESTORE,
+}	T_Action;
+
+/* Arguments needed for a worker process */
+typedef struct ParallelArgs
+{
+	struct _archiveHandle *AH;
+	struct _tocEntry *te;
+}	ParallelArgs;
+
+/* State for each parallel activity slot */
+typedef struct ParallelSlot
+{
+	ParallelArgs *args;
+	T_WorkerStatus workerStatus;
+	int			status;
+	int			pipeRead;
+	int			pipeWrite;
+	int			pipeRevRead;
+	int			pipeRevWrite;
+#ifdef WIN32
+	uintptr_t	hThread;
+	unsigned int threadId;
+#else
+	pid_t		pid;
+#endif
+} ParallelSlot;
+
+#define NO_SLOT (-1)
+
+typedef struct ParallelState
+{
+	int			numWorkers;
+	ParallelSlot *parallelSlot;
+} ParallelState;
+
+extern int	GetIdleWorker(ParallelState *pstate);
+extern bool IsEveryWorkerIdle(ParallelState *pstate);
+extern void ListenToWorkers(struct _archiveHandle * AH, ParallelState *pstate, bool do_wait);
+extern int	ReapWorkerStatus(ParallelState *pstate, int *status);
+extern void EnsureIdleWorker(struct _archiveHandle * AH, ParallelState *pstate);
+extern void EnsureWorkersFinished(struct _archiveHandle * AH, ParallelState *pstate);
+
+extern ParallelState *ParallelBackupStart(struct _archiveHandle * AH,
+					RestoreOptions *ropt);
+extern void DispatchJobForTocEntry(struct _archiveHandle * AH,
+					   ParallelState *pstate,
+					   struct _tocEntry * te, T_Action act);
+extern void ParallelBackupEnd(struct _archiveHandle * AH, ParallelState *pstate);
+
+extern void checkAborting(struct _archiveHandle * AH);

src/bin/pg_dump/pg_backup.h

@@ -82,9 +82,14 @@ struct Archive
 	int			minRemoteVersion;		/* allowable range */
 	int			maxRemoteVersion;
 
+	int			numWorkers;		/* number of parallel processes */
+	char	   *sync_snapshot_id;		/* sync snapshot id for parallel
+										 * operation */
+
 	/* info needed for string escaping */
 	int			encoding;		/* libpq code for client_encoding */
 	bool		std_strings;	/* standard_conforming_strings */
+	char	   *use_role;		/* Issue SET ROLE to this */
 
 	/* error handling */
 	bool		exit_on_error;	/* whether to exit on SQL errors... */
@@ -142,11 +147,12 @@ typedef struct _restoreOptions
 	int			suppressDumpWarnings;	/* Suppress output of WARNING entries
 										 * to stderr */
 	bool		single_txn;
-	int			number_of_jobs;
 
 	bool	   *idWanted;		/* array showing which dump IDs to emit */
 } RestoreOptions;
 
+typedef void (*SetupWorkerPtr) (Archive *AH, RestoreOptions *ropt);
+
 /*
  * Main archiver interface.
  */
@@ -189,7 +195,8 @@ extern Archive *OpenArchive(const char *FileSpec, const ArchiveFormat fmt);
 
 /* Create a new archive */
 extern Archive *CreateArchive(const char *FileSpec, const ArchiveFormat fmt,
-			  const int compression, ArchiveMode mode);
+			  const int compression, ArchiveMode mode,
+			  SetupWorkerPtr setupDumpWorker);
 
 /* The --list option */
 extern void PrintTOCSummary(Archive *AH, RestoreOptions *ropt);

src/bin/pg_dump/pg_backup_archiver.h

@@ -100,8 +100,21 @@ typedef z_stream *z_streamp;
 #define K_OFFSET_POS_SET 2
 #define K_OFFSET_NO_DATA 3
 
+/*
+ * Special exit values from worker children.  We reserve 0 for normal
+ * success; 1 and other small values should be interpreted as crashes.
+ */
+#define WORKER_OK					  0
+#define WORKER_CREATE_DONE			  10
+#define WORKER_INHIBIT_DATA			  11
+#define WORKER_IGNORED_ERRORS		  12
+
 struct _archiveHandle;
 struct _tocEntry;
+struct _restoreList;
+struct ParallelArgs;
+struct ParallelState;
+enum T_Action;
 
 typedef void (*ClosePtr) (struct _archiveHandle * AH);
 typedef void (*ReopenPtr) (struct _archiveHandle * AH);
@@ -129,6 +142,13 @@ typedef void (*PrintTocDataPtr) (struct _archiveHandle * AH, struct _tocEntry *
 typedef void (*ClonePtr) (struct _archiveHandle * AH);
 typedef void (*DeClonePtr) (struct _archiveHandle * AH);
 
+typedef char *(*WorkerJobRestorePtr) (struct _archiveHandle * AH, struct _tocEntry * te);
+typedef char *(*WorkerJobDumpPtr) (struct _archiveHandle * AH, struct _tocEntry * te);
+typedef char *(*MasterStartParallelItemPtr) (struct _archiveHandle * AH, struct _tocEntry * te,
+														 enum T_Action act);
+typedef int (*MasterEndParallelItemPtr) (struct _archiveHandle * AH, struct _tocEntry * te,
+										 const char *str, enum T_Action act);
+
 typedef size_t (*CustomOutPtr) (struct _archiveHandle * AH, const void *buf, size_t len);
 
 typedef enum
@@ -227,6 +247,13 @@ typedef struct _archiveHandle
 	StartBlobPtr StartBlobPtr;
 	EndBlobPtr EndBlobPtr;
 
+	MasterStartParallelItemPtr MasterStartParallelItemPtr;
+	MasterEndParallelItemPtr MasterEndParallelItemPtr;
+
+	SetupWorkerPtr SetupWorkerPtr;
+	WorkerJobDumpPtr WorkerJobDumpPtr;
+	WorkerJobRestorePtr WorkerJobRestorePtr;
+
 	ClonePtr ClonePtr;			/* Clone format-specific fields */
 	DeClonePtr DeClonePtr;		/* Clean up cloned fields */
 
@@ -236,6 +263,7 @@ typedef struct _archiveHandle
 	char	   *archdbname;		/* DB name *read* from archive */
 	enum trivalue promptPassword;
 	char	   *savedPassword;	/* password for ropt->username, if known */
+	char	   *use_role;
 	PGconn	   *connection;
 	int			connectToDB;	/* Flag to indicate if direct DB connection is
 								 * required */
@@ -327,6 +355,7 @@ typedef struct _tocEntry
 	int			nLockDeps;		/* number of such dependencies */
 } TocEntry;
 
+extern int	parallel_restore(struct ParallelArgs * args);
 extern void on_exit_close_archive(Archive *AHX);
 
 extern void warn_or_exit_horribly(ArchiveHandle *AH, const char *modulename, const char *fmt,...) __attribute__((format(PG_PRINTF_ATTRIBUTE, 3, 4)));
@@ -337,9 +366,13 @@ extern void WriteHead(ArchiveHandle *AH);
 extern void ReadHead(ArchiveHandle *AH);
 extern void WriteToc(ArchiveHandle *AH);
 extern void ReadToc(ArchiveHandle *AH);
-extern void WriteDataChunks(ArchiveHandle *AH);
+extern void WriteDataChunks(ArchiveHandle *AH, struct ParallelState *pstate);
+extern void WriteDataChunksForTocEntry(ArchiveHandle *AH, TocEntry *te);
+extern ArchiveHandle *CloneArchive(ArchiveHandle *AH);
+extern void DeCloneArchive(ArchiveHandle *AH);
 
 extern teReqs TocIDRequired(ArchiveHandle *AH, DumpId id);
+TocEntry   *getTocEntryByDumpId(ArchiveHandle *AH, DumpId id);
 extern bool checkSeek(FILE *fp);
 
 #define appendStringLiteralAHX(buf,str,AH) \

src/bin/pg_dump/pg_backup_custom.c

@@ -26,6 +26,7 @@
 
 #include "compress_io.h"
 #include "dumputils.h"
+#include "parallel.h"
 
 /*--------
  * Routines in the format interface
@@ -59,6 +60,10 @@ static void _LoadBlobs(ArchiveHandle *AH, bool drop);
 static void _Clone(ArchiveHandle *AH);
 static void _DeClone(ArchiveHandle *AH);
 
+static char *_MasterStartParallelItem(ArchiveHandle *AH, TocEntry *te, T_Action act);
+static int	_MasterEndParallelItem(ArchiveHandle *AH, TocEntry *te, const char *str, T_Action act);
+char	   *_WorkerJobRestoreCustom(ArchiveHandle *AH, TocEntry *te);
+
 typedef struct
 {
 	CompressorState *cs;
@@ -127,6 +132,13 @@ InitArchiveFmt_Custom(ArchiveHandle *AH)
 	AH->ClonePtr = _Clone;
 	AH->DeClonePtr = _DeClone;
 
+	AH->MasterStartParallelItemPtr = _MasterStartParallelItem;
+	AH->MasterEndParallelItemPtr = _MasterEndParallelItem;
+
+	/* no parallel dump in the custom archive, only parallel restore */
+	AH->WorkerJobDumpPtr = NULL;
+	AH->WorkerJobRestorePtr = _WorkerJobRestoreCustom;
+
 	/* Set up a private area. */
 	ctx = (lclContext *) pg_malloc0(sizeof(lclContext));
 	AH->formatData = (void *) ctx;
@@ -698,7 +710,7 @@ _CloseArchive(ArchiveHandle *AH)
 		tpos = ftello(AH->FH);
 		WriteToc(AH);
 		ctx->dataStart = _getFilePos(AH, ctx);
-		WriteDataChunks(AH);
+		WriteDataChunks(AH, NULL);
 
 		/*
 		 * If possible, re-write the TOC in order to update the data offset
@@ -796,6 +808,80 @@ _DeClone(ArchiveHandle *AH)
 	free(ctx);
 }
 
+/*
+ * This function is executed in the child of a parallel backup for the
+ * custom format archive and dumps the actual data.
+ */
+char *
+_WorkerJobRestoreCustom(ArchiveHandle *AH, TocEntry *te)
+{
+	/*
+	 * short fixed-size string + some ID so far, this needs to be malloc'ed
+	 * instead of static because we work with threads on windows
+	 */
+	const int	buflen = 64;
+	char	   *buf = (char *) pg_malloc(buflen);
+	ParallelArgs pargs;
+	int			status;
+
+	pargs.AH = AH;
+	pargs.te = te;
+
+	status = parallel_restore(&pargs);
+
+	snprintf(buf, buflen, "OK RESTORE %d %d %d", te->dumpId, status,
+			 status == WORKER_IGNORED_ERRORS ? AH->public.n_errors : 0);
+
+	return buf;
+}
+
+/*
+ * This function is executed in the parent process. Depending on the desired
+ * action (dump or restore) it creates a string that is understood by the
+ * _WorkerJobDump /_WorkerJobRestore functions of the dump format.
+ */
+static char *
+_MasterStartParallelItem(ArchiveHandle *AH, TocEntry *te, T_Action act)
+{
+	/*
+	 * A static char is okay here, even on Windows because we call this
+	 * function only from one process (the master).
+	 */
+	static char buf[64];		/* short fixed-size string + number */
+
+	/* no parallel dump in the custom archive format */
+	Assert(act == ACT_RESTORE);
+
+	snprintf(buf, sizeof(buf), "RESTORE %d", te->dumpId);
+
+	return buf;
+}
+
+/*
+ * This function is executed in the parent process. It analyzes the response of
+ * the _WorkerJobDump / _WorkerJobRestore functions of the dump format.
+ */
+static int
+_MasterEndParallelItem(ArchiveHandle *AH, TocEntry *te, const char *str, T_Action act)
+{
+	DumpId		dumpId;
+	int			nBytes,
+				status,
+				n_errors;
+
+	/* no parallel dump in the custom archive */
+	Assert(act == ACT_RESTORE);
+
+	sscanf(str, "%u %u %u%n", &dumpId, &status, &n_errors, &nBytes);
+
+	Assert(nBytes == strlen(str));
+	Assert(dumpId == te->dumpId);
+
+	AH->public.n_errors += n_errors;
+
+	return status;
+}
+
 /*--------------------------------------------------
  * END OF FORMAT CALLBACKS
  *--------------------------------------------------

src/bin/pg_dump/pg_backup_db.c

@@ -309,12 +309,30 @@ ConnectDatabase(Archive *AHX,
 	PQsetNoticeProcessor(AH->connection, notice_processor, NULL);
 }
 
+/*
+ * Close the connection to the database and also cancel off the query if we
+ * have one running.
+ */
 void
 DisconnectDatabase(Archive *AHX)
 {
 	ArchiveHandle *AH = (ArchiveHandle *) AHX;
+	PGcancel   *cancel;
+	char		errbuf[1];
 
-	PQfinish(AH->connection);	/* noop if AH->connection is NULL */
+	if (!AH->connection)
+		return;
+
+	if (PQtransactionStatus(AH->connection) == PQTRANS_ACTIVE)
+	{
+		if ((cancel = PQgetCancel(AH->connection)))
+		{
+			PQcancel(cancel, errbuf, sizeof(errbuf));
+			PQfreeCancel(cancel);
+		}
+	}
+
+	PQfinish(AH->connection);
 	AH->connection = NULL;
 }
 

src/bin/pg_dump/pg_backup_directory.c

@@ -35,6 +35,7 @@
 
 #include "compress_io.h"
 #include "dumputils.h"
+#include "parallel.h"
 
 #include <dirent.h>
 #include <sys/stat.h>
@@ -50,6 +51,7 @@ typedef struct
 	cfp		   *dataFH;			/* currently open data file */
 
 	cfp		   *blobsTocFH;		/* file handle for blobs.toc */
+	ParallelState *pstate;		/* for parallel backup / restore */
 } lclContext;
 
 typedef struct
@@ -70,6 +72,7 @@ static int	_ReadByte(ArchiveHandle *);
 static size_t _WriteBuf(ArchiveHandle *AH, const void *buf, size_t len);
 static size_t _ReadBuf(ArchiveHandle *AH, void *buf, size_t len);
 static void _CloseArchive(ArchiveHandle *AH);
+static void _ReopenArchive(ArchiveHandle *AH);
 static void _PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt);
 
 static void _WriteExtraToc(ArchiveHandle *AH, TocEntry *te);
@@ -82,8 +85,17 @@ static void _EndBlob(ArchiveHandle *AH, TocEntry *te, Oid oid);
 static void _EndBlobs(ArchiveHandle *AH, TocEntry *te);
 static void _LoadBlobs(ArchiveHandle *AH, RestoreOptions *ropt);
 
-static char *prependDirectory(ArchiveHandle *AH, const char *relativeFilename);
+static void _Clone(ArchiveHandle *AH);
+static void _DeClone(ArchiveHandle *AH);
 
+static char *_MasterStartParallelItem(ArchiveHandle *AH, TocEntry *te, T_Action act);
+static int _MasterEndParallelItem(ArchiveHandle *AH, TocEntry *te,
+					   const char *str, T_Action act);
+static char *_WorkerJobRestoreDirectory(ArchiveHandle *AH, TocEntry *te);
+static char *_WorkerJobDumpDirectory(ArchiveHandle *AH, TocEntry *te);
+
+static void setFilePath(ArchiveHandle *AH, char *buf,
+			const char *relativeFilename);
 
 /*
  *	Init routine required by ALL formats. This is a global routine
@@ -110,7 +122,7 @@ InitArchiveFmt_Directory(ArchiveHandle *AH)
 	AH->WriteBufPtr = _WriteBuf;
 	AH->ReadBufPtr = _ReadBuf;
 	AH->ClosePtr = _CloseArchive;
-	AH->ReopenPtr = NULL;
+	AH->ReopenPtr = _ReopenArchive;
 	AH->PrintTocDataPtr = _PrintTocData;
 	AH->ReadExtraTocPtr = _ReadExtraToc;
 	AH->WriteExtraTocPtr = _WriteExtraToc;
@@ -121,8 +133,14 @@ InitArchiveFmt_Directory(ArchiveHandle *AH)
 	AH->EndBlobPtr = _EndBlob;
 	AH->EndBlobsPtr = _EndBlobs;
 
-	AH->ClonePtr = NULL;
+	AH->ClonePtr = _Clone;
-	AH->DeClonePtr = NULL;
+	AH->DeClonePtr = _DeClone;
+
+	AH->WorkerJobRestorePtr = _WorkerJobRestoreDirectory;
+	AH->WorkerJobDumpPtr = _WorkerJobDumpDirectory;
+
+	AH->MasterStartParallelItemPtr = _MasterStartParallelItem;
+	AH->MasterEndParallelItemPtr = _MasterEndParallelItem;
 
 	/* Set up our private context */
 	ctx = (lclContext *) pg_malloc0(sizeof(lclContext));
@@ -146,16 +164,41 @@ InitArchiveFmt_Directory(ArchiveHandle *AH)
 
 	if (AH->mode == archModeWrite)
 	{
-		if (mkdir(ctx->directory, 0700) < 0)
+		struct stat st;
+		bool		is_empty = false;
+
+		/* we accept an empty existing directory */
+		if (stat(ctx->directory, &st) == 0 && S_ISDIR(st.st_mode))
+		{
+			DIR		   *dir = opendir(ctx->directory);
+
+			if (dir)
+			{
+				struct dirent *d;
+
+				is_empty = true;
+				while ((d = readdir(dir)))
+				{
+					if (strcmp(d->d_name, ".") != 0 && strcmp(d->d_name, "..") != 0)
+					{
+						is_empty = false;
+						break;
+					}
+				}
+				closedir(dir);
+			}
+		}
+
+		if (!is_empty && mkdir(ctx->directory, 0700) < 0)
 			exit_horribly(modulename, "could not create directory \"%s\": %s\n",
 						  ctx->directory, strerror(errno));
 	}
 	else
 	{							/* Read Mode */
-		char	   *fname;
+		char		fname[MAXPGPATH];
 		cfp		   *tocFH;
 
-		fname = prependDirectory(AH, "toc.dat");
+		setFilePath(AH, fname, "toc.dat");
 
 		tocFH = cfopen_read(fname, PG_BINARY_R);
 		if (tocFH == NULL)
@@ -281,9 +324,9 @@ _StartData(ArchiveHandle *AH, TocEntry *te)
 {
 	lclTocEntry *tctx = (lclTocEntry *) te->formatData;
 	lclContext *ctx = (lclContext *) AH->formatData;
-	char	   *fname;
+	char		fname[MAXPGPATH];
 
-	fname = prependDirectory(AH, tctx->filename);
+	setFilePath(AH, fname, tctx->filename);
 
 	ctx->dataFH = cfopen_write(fname, PG_BINARY_W, AH->compression);
 	if (ctx->dataFH == NULL)
@@ -308,6 +351,9 @@ _WriteData(ArchiveHandle *AH, const void *data, size_t dLen)
 	if (dLen == 0)
 		return 0;
 
+	/* Are we aborting? */
+	checkAborting(AH);
+
 	return cfwrite(data, dLen, ctx->dataFH);
 }
 
@@ -375,8 +421,9 @@ _PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt)
 		_LoadBlobs(AH, ropt);
 	else
 	{
-		char	   *fname = prependDirectory(AH, tctx->filename);
+		char		fname[MAXPGPATH];
 
+		setFilePath(AH, fname, tctx->filename);
 		_PrintFileData(AH, fname, ropt);
 	}
 }
@@ -386,12 +433,12 @@ _LoadBlobs(ArchiveHandle *AH, RestoreOptions *ropt)
 {
 	Oid			oid;
 	lclContext *ctx = (lclContext *) AH->formatData;
-	char	   *fname;
+	char		fname[MAXPGPATH];
 	char		line[MAXPGPATH];
 
 	StartRestoreBlobs(AH);
 
-	fname = prependDirectory(AH, "blobs.toc");
+	setFilePath(AH, fname, "blobs.toc");
 
 	ctx->blobsTocFH = cfopen_read(fname, PG_BINARY_R);
 
@@ -474,6 +521,9 @@ _WriteBuf(ArchiveHandle *AH, const void *buf, size_t len)
 	lclContext *ctx = (lclContext *) AH->formatData;
 	size_t		res;
 
+	/* Are we aborting? */
+	checkAborting(AH);
+
 	res = cfwrite(buf, len, ctx->dataFH);
 	if (res != len)
 		exit_horribly(modulename, "could not write to output file: %s\n",
@@ -518,7 +568,12 @@ _CloseArchive(ArchiveHandle *AH)
 	if (AH->mode == archModeWrite)
 	{
 		cfp		   *tocFH;
-		char	   *fname = prependDirectory(AH, "toc.dat");
+		char		fname[MAXPGPATH];
+
+		setFilePath(AH, fname, "toc.dat");
+
+		/* this will actually fork the processes for a parallel backup */
+		ctx->pstate = ParallelBackupStart(AH, NULL);
 
 		/* The TOC is always created uncompressed */
 		tocFH = cfopen_write(fname, PG_BINARY_W, 0);
@@ -539,11 +594,25 @@ _CloseArchive(ArchiveHandle *AH)
 		if (cfclose(tocFH) != 0)
 			exit_horribly(modulename, "could not close TOC file: %s\n",
 						  strerror(errno));
-		WriteDataChunks(AH);
+		WriteDataChunks(AH, ctx->pstate);
+
+		ParallelBackupEnd(AH, ctx->pstate);
 	}
 	AH->FH = NULL;
 }
 
+/*
+ * Reopen the archive's file handle.
+ */
+static void
+_ReopenArchive(ArchiveHandle *AH)
+{
+	/*
+	 * Our TOC is in memory, our data files are opened by each child anyway as
+	 * they are separate. We support reopening the archive by just doing
+	 * nothing.
+	 */
+}
 
 /*
  * BLOB support
@@ -560,9 +629,9 @@ static void
 _StartBlobs(ArchiveHandle *AH, TocEntry *te)
 {
 	lclContext *ctx = (lclContext *) AH->formatData;
-	char	   *fname;
+	char		fname[MAXPGPATH];
 
-	fname = prependDirectory(AH, "blobs.toc");
+	setFilePath(AH, fname, "blobs.toc");
 
 	/* The blob TOC file is never compressed */
 	ctx->blobsTocFH = cfopen_write(fname, "ab", 0);
@@ -627,12 +696,16 @@ _EndBlobs(ArchiveHandle *AH, TocEntry *te)
 	ctx->blobsTocFH = NULL;
 }
 
-
+/*
-static char *
+ * Gets a relative file name and prepends the output directory, writing the
-prependDirectory(ArchiveHandle *AH, const char *relativeFilename)
+ * result to buf. The caller needs to make sure that buf is MAXPGPATH bytes
+ * big. Can't use a static char[MAXPGPATH] inside the function because we run
+ * multithreaded on Windows.
+ */
+static void
+setFilePath(ArchiveHandle *AH, char *buf, const char *relativeFilename)
 {
 	lclContext *ctx = (lclContext *) AH->formatData;
-	static char buf[MAXPGPATH];
 	char	   *dname;
 
 	dname = ctx->directory;
@@ -643,6 +716,157 @@ prependDirectory(ArchiveHandle *AH, const char *relativeFilename)
 	strcpy(buf, dname);
 	strcat(buf, "/");
 	strcat(buf, relativeFilename);
+}
+
+/*
+ * Clone format-specific fields during parallel restoration.
+ */
+static void
+_Clone(ArchiveHandle *AH)
+{
+	lclContext *ctx = (lclContext *) AH->formatData;
+
+	AH->formatData = (lclContext *) pg_malloc(sizeof(lclContext));
+	memcpy(AH->formatData, ctx, sizeof(lclContext));
+	ctx = (lclContext *) AH->formatData;
+
+	/*
+	 * Note: we do not make a local lo_buf because we expect at most one BLOBS
+	 * entry per archive, so no parallelism is possible.  Likewise,
+	 * TOC-entry-local state isn't an issue because any one TOC entry is
+	 * touched by just one worker child.
+	 */
+
+	/*
+	 * We also don't copy the ParallelState pointer (pstate), only the master
+	 * process ever writes to it.
+	 */
+}
+
+static void
+_DeClone(ArchiveHandle *AH)
+{
+	lclContext *ctx = (lclContext *) AH->formatData;
+
+	free(ctx);
+}
+
+/*
+ * This function is executed in the parent process. Depending on the desired
+ * action (dump or restore) it creates a string that is understood by the
+ * _WorkerJobDump /_WorkerJobRestore functions of the dump format.
+ */
+static char *
+_MasterStartParallelItem(ArchiveHandle *AH, TocEntry *te, T_Action act)
+{
+	/*
+	 * A static char is okay here, even on Windows because we call this
+	 * function only from one process (the master).
+	 */
+	static char buf[64];
+
+	if (act == ACT_DUMP)
+		snprintf(buf, sizeof(buf), "DUMP %d", te->dumpId);
+	else if (act == ACT_RESTORE)
+		snprintf(buf, sizeof(buf), "RESTORE %d", te->dumpId);
 
 	return buf;
 }
+
+/*
+ * This function is executed in the child of a parallel backup for the
+ * directory archive and dumps the actual data.
+ *
+ * We are currently returning only the DumpId so theoretically we could
+ * make this function returning an int (or a DumpId). However, to
+ * facilitate further enhancements and because sooner or later we need to
+ * convert this to a string and send it via a message anyway, we stick with
+ * char *. It is parsed on the other side by the _EndMasterParallel()
+ * function of the respective dump format.
+ */
+static char *
+_WorkerJobDumpDirectory(ArchiveHandle *AH, TocEntry *te)
+{
+	/*
+	 * short fixed-size string + some ID so far, this needs to be malloc'ed
+	 * instead of static because we work with threads on windows
+	 */
+	const int	buflen = 64;
+	char	   *buf = (char *) pg_malloc(buflen);
+	lclTocEntry *tctx = (lclTocEntry *) te->formatData;
+
+	/* This should never happen */
+	if (!tctx)
+		exit_horribly(modulename, "Error during backup\n");
+
+	/*
+	 * This function returns void. We either fail and die horribly or
+	 * succeed... A failure will be detected by the parent when the child dies
+	 * unexpectedly.
+	 */
+	WriteDataChunksForTocEntry(AH, te);
+
+	snprintf(buf, buflen, "OK DUMP %d", te->dumpId);
+
+	return buf;
+}
+
+/*
+ * This function is executed in the child of a parallel backup for the
+ * directory archive and dumps the actual data.
+ */
+static char *
+_WorkerJobRestoreDirectory(ArchiveHandle *AH, TocEntry *te)
+{
+	/*
+	 * short fixed-size string + some ID so far, this needs to be malloc'ed
+	 * instead of static because we work with threads on windows
+	 */
+	const int	buflen = 64;
+	char	   *buf = (char *) pg_malloc(buflen);
+	ParallelArgs pargs;
+	int			status;
+
+	pargs.AH = AH;
+	pargs.te = te;
+
+	status = parallel_restore(&pargs);
+
+	snprintf(buf, buflen, "OK RESTORE %d %d %d", te->dumpId, status,
+			 status == WORKER_IGNORED_ERRORS ? AH->public.n_errors : 0);
+
+	return buf;
+}
+
+/*
+ * This function is executed in the parent process. It analyzes the response of
+ * the _WorkerJobDumpDirectory/_WorkerJobRestoreDirectory functions of the
+ * respective dump format.
+ */
+static int
+_MasterEndParallelItem(ArchiveHandle *AH, TocEntry *te, const char *str, T_Action act)
+{
+	DumpId		dumpId;
+	int			nBytes,
+				n_errors;
+	int			status = 0;
+
+	if (act == ACT_DUMP)
+	{
+		sscanf(str, "%u%n", &dumpId, &nBytes);
+
+		Assert(dumpId == te->dumpId);
+		Assert(nBytes == strlen(str));
+	}
+	else if (act == ACT_RESTORE)
+	{
+		sscanf(str, "%u %u %u%n", &dumpId, &status, &n_errors, &nBytes);
+
+		Assert(dumpId == te->dumpId);
+		Assert(nBytes == strlen(str));
+
+		AH->public.n_errors += n_errors;
+	}
+
+	return status;
+}

src/bin/pg_dump/pg_backup_tar.c

@@ -158,6 +158,12 @@ InitArchiveFmt_Tar(ArchiveHandle *AH)
 	AH->ClonePtr = NULL;
 	AH->DeClonePtr = NULL;
 
+	AH->MasterStartParallelItemPtr = NULL;
+	AH->MasterEndParallelItemPtr = NULL;
+
+	AH->WorkerJobDumpPtr = NULL;
+	AH->WorkerJobRestorePtr = NULL;
+
 	/*
 	 * Set up some special context used in compressing data.
 	 */
@@ -828,7 +834,7 @@ _CloseArchive(ArchiveHandle *AH)
 		/*
 		 * Now send the data (tables & blobs)
 		 */
-		WriteDataChunks(AH);
+		WriteDataChunks(AH, NULL);
 
 		/*
 		 * Now this format wants to append a script which does a full restore

src/bin/pg_dump/pg_dump.c

@@ -135,6 +135,7 @@ static int	disable_dollar_quoting = 0;
 static int	dump_inserts = 0;
 static int	column_inserts = 0;
 static int	no_security_labels = 0;
+static int	no_synchronized_snapshots = 0;
 static int	no_unlogged_table_data = 0;
 static int	serializable_deferrable = 0;
 
@@ -243,8 +244,6 @@ static Oid	findLastBuiltinOid_V70(Archive *fout);
 static void selectSourceSchema(Archive *fout, const char *schemaName);
 static char *getFormattedTypeName(Archive *fout, Oid oid, OidOptions opts);
 static char *myFormatType(const char *typname, int32 typmod);
-static const char *fmtQualifiedId(Archive *fout,
-			   const char *schema, const char *id);
 static void getBlobs(Archive *fout);
 static void dumpBlob(Archive *fout, BlobInfo *binfo);
 static int	dumpBlobs(Archive *fout, void *arg);
@@ -262,8 +261,10 @@ static void binary_upgrade_extension_member(PQExpBuffer upgrade_buffer,
 								DumpableObject *dobj,
 								const char *objlabel);
 static const char *getAttrName(int attrnum, TableInfo *tblInfo);
-static const char *fmtCopyColumnList(const TableInfo *ti);
+static const char *fmtCopyColumnList(const TableInfo *ti, PQExpBuffer buffer);
+static char *get_synchronized_snapshot(Archive *fout);
 static PGresult *ExecuteSqlQueryForSingleRow(Archive *fout, char *query);
+static void setupDumpWorker(Archive *AHX, RestoreOptions *ropt);
 
 
 int
@@ -284,6 +285,7 @@ main(int argc, char **argv)
 	int			numObjs;
 	DumpableObject *boundaryObjs;
 	int			i;
+	int			numWorkers = 1;
 	enum trivalue prompt_password = TRI_DEFAULT;
 	int			compressLevel = -1;
 	int			plainText = 0;
@@ -314,6 +316,7 @@ main(int argc, char **argv)
 		{"format", required_argument, NULL, 'F'},
 		{"host", required_argument, NULL, 'h'},
 		{"ignore-version", no_argument, NULL, 'i'},
+		{"jobs", 1, NULL, 'j'},
 		{"no-reconnect", no_argument, NULL, 'R'},
 		{"oids", no_argument, NULL, 'o'},
 		{"no-owner", no_argument, NULL, 'O'},
@@ -353,6 +356,7 @@ main(int argc, char **argv)
 		{"serializable-deferrable", no_argument, &serializable_deferrable, 1},
 		{"use-set-session-authorization", no_argument, &use_setsessauth, 1},
 		{"no-security-labels", no_argument, &no_security_labels, 1},
+		{"no-synchronized-snapshots", no_argument, &no_synchronized_snapshots, 1},
 		{"no-unlogged-table-data", no_argument, &no_unlogged_table_data, 1},
 
 		{NULL, 0, NULL, 0}
@@ -360,6 +364,12 @@ main(int argc, char **argv)
 
 	set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("pg_dump"));
 
+	/*
+	 * Initialize what we need for parallel execution, especially for thread
+	 * support on Windows.
+	 */
+	init_parallel_dump_utils();
+
 	g_verbose = false;
 
 	strcpy(g_comment_start, "-- ");
@@ -390,7 +400,7 @@ main(int argc, char **argv)
 		}
 	}
 
-	while ((c = getopt_long(argc, argv, "abcCd:E:f:F:h:iK:n:N:oOp:RsS:t:T:U:vwWxZ:",
+	while ((c = getopt_long(argc, argv, "abcCd:E:f:F:h:ij:K:n:N:oOp:RsS:t:T:U:vwWxZ:",
 							long_options, &optindex)) != -1)
 	{
 		switch (c)
@@ -435,6 +445,10 @@ main(int argc, char **argv)
 				/* ignored, deprecated option */
 				break;
 
+			case 'j':			/* number of dump jobs */
+				numWorkers = atoi(optarg);
+				break;
+
 			case 'n':			/* include schema(s) */
 				simple_string_list_append(&schema_include_patterns, optarg);
 				include_everything = false;
@@ -577,8 +591,25 @@ main(int argc, char **argv)
 			compressLevel = 0;
 	}
 
+	/*
+	 * On Windows we can only have at most MAXIMUM_WAIT_OBJECTS (= 64 usually)
+	 * parallel jobs because that's the maximum limit for the
+	 * WaitForMultipleObjects() call.
+	 */
+	if (numWorkers <= 0
+#ifdef WIN32
+		|| numWorkers > MAXIMUM_WAIT_OBJECTS
+#endif
+		)
+		exit_horribly(NULL, "%s: invalid number of parallel jobs\n", progname);
+
+	/* Parallel backup only in the directory archive format so far */
+	if (archiveFormat != archDirectory && numWorkers > 1)
+		exit_horribly(NULL, "parallel backup only supported by the directory format\n");
+
 	/* Open the output file */
-	fout = CreateArchive(filename, archiveFormat, compressLevel, archiveMode);
+	fout = CreateArchive(filename, archiveFormat, compressLevel, archiveMode,
+						 setupDumpWorker);
 
 	/* Register the cleanup hook */
 	on_exit_close_archive(fout);
@@ -600,6 +631,8 @@ main(int argc, char **argv)
 	fout->minRemoteVersion = 70000;
 	fout->maxRemoteVersion = (my_version / 100) * 100 + 99;
 
+	fout->numWorkers = numWorkers;
+
 	/*
 	 * Open the database using the Archiver, so it knows about it. Errors mean
 	 * death.
@@ -621,6 +654,7 @@ main(int argc, char **argv)
 	if (fout->remoteVersion >= 90000)
 	{
 		PGresult   *res = ExecuteSqlQueryForSingleRow(fout, "SELECT pg_catalog.pg_is_in_recovery()");
+
 		if (strcmp(PQgetvalue(res, 0, 0), "t") == 0)
 		{
 			/*
@@ -632,32 +666,6 @@ main(int argc, char **argv)
 		PQclear(res);
 	}
 
-	/*
-	 * Start transaction-snapshot mode transaction to dump consistent data.
-	 */
-	ExecuteSqlStatement(fout, "BEGIN");
-	if (fout->remoteVersion >= 90100)
-	{
-		if (serializable_deferrable)
-			ExecuteSqlStatement(fout,
-								"SET TRANSACTION ISOLATION LEVEL "
-								"SERIALIZABLE, READ ONLY, DEFERRABLE");
-		else
-			ExecuteSqlStatement(fout,
-								"SET TRANSACTION ISOLATION LEVEL "
-								"REPEATABLE READ, READ ONLY");
-	}
-	else if (fout->remoteVersion >= 70400)
-	{
-		/* note: comma was not accepted in SET TRANSACTION before 8.0 */
-		ExecuteSqlStatement(fout,
-							"SET TRANSACTION ISOLATION LEVEL "
-							"SERIALIZABLE READ ONLY");
-	}
-	else
-		ExecuteSqlStatement(fout,
-							"SET TRANSACTION ISOLATION LEVEL SERIALIZABLE");
-
 	/* Select the appropriate subquery to convert user IDs to names */
 	if (fout->remoteVersion >= 80100)
 		username_subquery = "SELECT rolname FROM pg_catalog.pg_roles WHERE oid =";
@@ -666,6 +674,14 @@ main(int argc, char **argv)
 	else
 		username_subquery = "SELECT usename FROM pg_user WHERE usesysid =";
 
+	/* check the version for the synchronized snapshots feature */
+	if (numWorkers > 1 && fout->remoteVersion < 90200
+		&& !no_synchronized_snapshots)
+		exit_horribly(NULL,
+			  "No synchronized snapshots available in this server version.\n"
+			   "Run with --no-synchronized-snapshots instead if you do not\n"
+					  "need synchronized snapshots.\n");
+
 	/* Find the last built-in OID, if needed */
 	if (fout->remoteVersion < 70300)
 	{
@@ -763,6 +779,10 @@ main(int argc, char **argv)
 	else
 		sortDumpableObjectsByTypeOid(dobjs, numObjs);
 
+	/* If we do a parallel dump, we want the largest tables to go first */
+	if (archiveFormat == archDirectory && numWorkers > 1)
+		sortDataAndIndexObjectsBySize(dobjs, numObjs);
+
 	sortDumpableObjects(dobjs, numObjs,
 						boundaryObjs[0].dumpId, boundaryObjs[1].dumpId);
 
@@ -810,9 +830,9 @@ main(int argc, char **argv)
 	SetArchiveRestoreOptions(fout, ropt);
 
 	/*
-	 * The archive's TOC entries are now marked as to which ones will
+	 * The archive's TOC entries are now marked as to which ones will actually
-	 * actually be output, so we can set up their dependency lists properly.
+	 * be output, so we can set up their dependency lists properly. This isn't
-	 * This isn't necessary for plain-text output, though.
+	 * necessary for plain-text output, though.
 	 */
 	if (!plainText)
 		BuildArchiveDependencies(fout);
@@ -844,6 +864,7 @@ help(const char *progname)
 	printf(_("  -f, --file=FILENAME          output file or directory name\n"));
 	printf(_("  -F, --format=c|d|t|p         output file format (custom, directory, tar,\n"
 			 "                               plain text (default))\n"));
+	printf(_("  -j, --jobs=NUM               use this many parallel jobs to dump\n"));
 	printf(_("  -v, --verbose                verbose mode\n"));
 	printf(_("  -V, --version                output version information, then exit\n"));
 	printf(_("  -Z, --compress=0-9           compression level for compressed formats\n"));
@@ -873,6 +894,7 @@ help(const char *progname)
 	printf(_("  --exclude-table-data=TABLE   do NOT dump data for the named table(s)\n"));
 	printf(_("  --inserts                    dump data as INSERT commands, rather than COPY\n"));
 	printf(_("  --no-security-labels         do not dump security label assignments\n"));
+	printf(_("  --no-synchronized-snapshots parallel processes should not use synchronized snapshots\n"));
 	printf(_("  --no-tablespaces             do not dump tablespace assignments\n"));
 	printf(_("  --no-unlogged-table-data     do not dump unlogged table data\n"));
 	printf(_("  --quote-all-identifiers      quote all identifiers, even if not key words\n"));
@@ -902,7 +924,12 @@ setup_connection(Archive *AH, const char *dumpencoding, char *use_role)
 	PGconn	   *conn = GetConnection(AH);
 	const char *std_strings;
 
-	/* Set the client encoding if requested */
+	/*
+	 * Set the client encoding if requested. If dumpencoding == NULL then
+	 * either it hasn't been requested or we're a cloned connection and then
+	 * this has already been set in CloneArchive according to the original
+	 * connection encoding.
+	 */
 	if (dumpencoding)
 	{
 		if (PQsetClientEncoding(conn, dumpencoding) < 0)
@@ -919,6 +946,10 @@ setup_connection(Archive *AH, const char *dumpencoding, char *use_role)
 	std_strings = PQparameterStatus(conn, "standard_conforming_strings");
 	AH->std_strings = (std_strings && strcmp(std_strings, "on") == 0);
 
+	/* Set the role if requested */
+	if (!use_role && AH->use_role)
+		use_role = AH->use_role;
+
 	/* Set the role if requested */
 	if (use_role && AH->remoteVersion >= 80100)
 	{
@@ -927,6 +958,10 @@ setup_connection(Archive *AH, const char *dumpencoding, char *use_role)
 		appendPQExpBuffer(query, "SET ROLE %s", fmtId(use_role));
 		ExecuteSqlStatement(AH, query->data);
 		destroyPQExpBuffer(query);
+
+		/* save this for later use on parallel connections */
+		if (!AH->use_role)
+			AH->use_role = strdup(use_role);
 	}
 
 	/* Set the datestyle to ISO to ensure the dump's portability */
@@ -965,6 +1000,68 @@ setup_connection(Archive *AH, const char *dumpencoding, char *use_role)
 	 */
 	if (quote_all_identifiers && AH->remoteVersion >= 90100)
 		ExecuteSqlStatement(AH, "SET quote_all_identifiers = true");
+
+	/*
+	 * Start transaction-snapshot mode transaction to dump consistent data.
+	 */
+	ExecuteSqlStatement(AH, "BEGIN");
+	if (AH->remoteVersion >= 90100)
+	{
+		if (serializable_deferrable)
+			ExecuteSqlStatement(AH,
+								"SET TRANSACTION ISOLATION LEVEL "
+								"SERIALIZABLE, READ ONLY, DEFERRABLE");
+		else
+			ExecuteSqlStatement(AH,
+								"SET TRANSACTION ISOLATION LEVEL "
+								"REPEATABLE READ, READ ONLY");
+	}
+	else if (AH->remoteVersion >= 70400)
+	{
+		/* note: comma was not accepted in SET TRANSACTION before 8.0 */
+		ExecuteSqlStatement(AH,
+							"SET TRANSACTION ISOLATION LEVEL "
+							"SERIALIZABLE READ ONLY");
+	}
+	else
+		ExecuteSqlStatement(AH,
+							"SET TRANSACTION ISOLATION LEVEL SERIALIZABLE");
+
+
+
+	if (AH->numWorkers > 1 && AH->remoteVersion >= 90200 && !no_synchronized_snapshots)
+	{
+		if (AH->sync_snapshot_id)
+		{
+			PQExpBuffer query = createPQExpBuffer();
+
+			appendPQExpBuffer(query, "SET TRANSACTION SNAPSHOT ");
+			appendStringLiteralConn(query, AH->sync_snapshot_id, conn);
+			destroyPQExpBuffer(query);
+		}
+		else
+			AH->sync_snapshot_id = get_synchronized_snapshot(AH);
+	}
+}
+
+static void
+setupDumpWorker(Archive *AHX, RestoreOptions *ropt)
+{
+	setup_connection(AHX, NULL, NULL);
+}
+
+static char *
+get_synchronized_snapshot(Archive *fout)
+{
+	char	   *query = "SELECT pg_export_snapshot()";
+	char	   *result;
+	PGresult   *res;
+
+	res = ExecuteSqlQueryForSingleRow(fout, query);
+	result = strdup(PQgetvalue(res, 0, 0));
+	PQclear(res);
+
+	return result;
 }
 
 static ArchiveFormat
@@ -1282,6 +1379,12 @@ dumpTableData_copy(Archive *fout, void *dcontext)
 	const bool	hasoids = tbinfo->hasoids;
 	const bool	oids = tdinfo->oids;
 	PQExpBuffer q = createPQExpBuffer();
+
+	/*
+	 * Note: can't use getThreadLocalPQExpBuffer() here, we're calling fmtId
+	 * which uses it already.
+	 */
+	PQExpBuffer clistBuf = createPQExpBuffer();
 	PGconn	   *conn = GetConnection(fout);
 	PGresult   *res;
 	int			ret;
@@ -1306,14 +1409,14 @@ dumpTableData_copy(Archive *fout, void *dcontext)
 	 * cases involving ADD COLUMN and inheritance.)
 	 */
 	if (fout->remoteVersion >= 70300)
-		column_list = fmtCopyColumnList(tbinfo);
+		column_list = fmtCopyColumnList(tbinfo, clistBuf);
 	else
 		column_list = "";		/* can't select columns in COPY */
 
 	if (oids && hasoids)
 	{
 		appendPQExpBuffer(q, "COPY %s %s WITH OIDS TO stdout;",
-						  fmtQualifiedId(fout,
+						  fmtQualifiedId(fout->remoteVersion,
 										 tbinfo->dobj.namespace->dobj.name,
 										 classname),
 						  column_list);
@@ -1331,7 +1434,7 @@ dumpTableData_copy(Archive *fout, void *dcontext)
 		else
 			appendPQExpBufferStr(q, "* ");
 		appendPQExpBuffer(q, "FROM %s %s) TO stdout;",
-						  fmtQualifiedId(fout,
+						  fmtQualifiedId(fout->remoteVersion,
 										 tbinfo->dobj.namespace->dobj.name,
 										 classname),
 						  tdinfo->filtercond);
@@ -1339,13 +1442,14 @@ dumpTableData_copy(Archive *fout, void *dcontext)
 	else
 	{
 		appendPQExpBuffer(q, "COPY %s %s TO stdout;",
-						  fmtQualifiedId(fout,
+						  fmtQualifiedId(fout->remoteVersion,
 										 tbinfo->dobj.namespace->dobj.name,
 										 classname),
 						  column_list);
 	}
 	res = ExecuteSqlQuery(fout, q->data, PGRES_COPY_OUT);
 	PQclear(res);
+	destroyPQExpBuffer(clistBuf);
 
 	for (;;)
 	{
@@ -1464,7 +1568,7 @@ dumpTableData_insert(Archive *fout, void *dcontext)
 	{
 		appendPQExpBuffer(q, "DECLARE _pg_dump_cursor CURSOR FOR "
 						  "SELECT * FROM ONLY %s",
-						  fmtQualifiedId(fout,
+						  fmtQualifiedId(fout->remoteVersion,
 										 tbinfo->dobj.namespace->dobj.name,
 										 classname));
 	}
@@ -1472,7 +1576,7 @@ dumpTableData_insert(Archive *fout, void *dcontext)
 	{
 		appendPQExpBuffer(q, "DECLARE _pg_dump_cursor CURSOR FOR "
 						  "SELECT * FROM %s",
-						  fmtQualifiedId(fout,
+						  fmtQualifiedId(fout->remoteVersion,
 										 tbinfo->dobj.namespace->dobj.name,
 										 classname));
 	}
@@ -1604,6 +1708,7 @@ dumpTableData(Archive *fout, TableDataInfo *tdinfo)
 {
 	TableInfo  *tbinfo = tdinfo->tdtable;
 	PQExpBuffer copyBuf = createPQExpBuffer();
+	PQExpBuffer clistBuf = createPQExpBuffer();
 	DataDumperPtr dumpFn;
 	char	   *copyStmt;
 
@@ -1615,7 +1720,7 @@ dumpTableData(Archive *fout, TableDataInfo *tdinfo)
 		appendPQExpBuffer(copyBuf, "COPY %s ",
 						  fmtId(tbinfo->dobj.name));
 		appendPQExpBuffer(copyBuf, "%s %sFROM stdin;\n",
-						  fmtCopyColumnList(tbinfo),
+						  fmtCopyColumnList(tbinfo, clistBuf),
 					  (tdinfo->oids && tbinfo->hasoids) ? "WITH OIDS " : "");
 		copyStmt = copyBuf->data;
 	}
@@ -1640,6 +1745,7 @@ dumpTableData(Archive *fout, TableDataInfo *tdinfo)
 				 dumpFn, tdinfo);
 
 	destroyPQExpBuffer(copyBuf);
+	destroyPQExpBuffer(clistBuf);
 }
 
 /*
@@ -4122,6 +4228,7 @@ getTables(Archive *fout, int *numTables)
 	int			i_reloptions;
 	int			i_toastreloptions;
 	int			i_reloftype;
+	int			i_relpages;
 
 	/* Make sure we are in proper schema */
 	selectSourceSchema(fout, "pg_catalog");
@@ -4161,6 +4268,7 @@ getTables(Archive *fout, int *numTables)
 						  "c.relfrozenxid, tc.oid AS toid, "
 						  "tc.relfrozenxid AS tfrozenxid, "
 		 "c.relpersistence, pg_relation_is_scannable(c.oid) as isscannable, "
+						  "c.relpages, "
 						  "CASE WHEN c.reloftype <> 0 THEN c.reloftype::pg_catalog.regtype ELSE NULL END AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4233,6 +4341,7 @@ getTables(Archive *fout, int *numTables)
 						  "c.relfrozenxid, tc.oid AS toid, "
 						  "tc.relfrozenxid AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "c.relpages, "
 						  "CASE WHEN c.reloftype <> 0 THEN c.reloftype::pg_catalog.regtype ELSE NULL END AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4268,6 +4377,7 @@ getTables(Archive *fout, int *numTables)
 						  "c.relfrozenxid, tc.oid AS toid, "
 						  "tc.relfrozenxid AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "c.relpages, "
 						  "NULL AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4303,6 +4413,7 @@ getTables(Archive *fout, int *numTables)
 						  "c.relfrozenxid, tc.oid AS toid, "
 						  "tc.relfrozenxid AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "c.relpages, "
 						  "NULL AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4339,6 +4450,7 @@ getTables(Archive *fout, int *numTables)
 						  "0 AS toid, "
 						  "0 AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "relpages, "
 						  "NULL AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4374,6 +4486,7 @@ getTables(Archive *fout, int *numTables)
 						  "0 AS toid, "
 						  "0 AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "relpages, "
 						  "NULL AS reloftype, "
 						  "d.refobjid AS owning_tab, "
 						  "d.refobjsubid AS owning_col, "
@@ -4405,6 +4518,7 @@ getTables(Archive *fout, int *numTables)
 						  "0 AS toid, "
 						  "0 AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "relpages, "
 						  "NULL AS reloftype, "
 						  "NULL::oid AS owning_tab, "
 						  "NULL::int4 AS owning_col, "
@@ -4431,6 +4545,7 @@ getTables(Archive *fout, int *numTables)
 						  "0 AS toid, "
 						  "0 AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "relpages, "
 						  "NULL AS reloftype, "
 						  "NULL::oid AS owning_tab, "
 						  "NULL::int4 AS owning_col, "
@@ -4467,6 +4582,7 @@ getTables(Archive *fout, int *numTables)
 						  "0 AS toid, "
 						  "0 AS tfrozenxid, "
 						  "'p' AS relpersistence, 't'::bool as isscannable, "
+						  "0 AS relpages, "
 						  "NULL AS reloftype, "
 						  "NULL::oid AS owning_tab, "
 						  "NULL::int4 AS owning_col, "
@@ -4515,6 +4631,7 @@ getTables(Archive *fout, int *numTables)
 	i_toastfrozenxid = PQfnumber(res, "tfrozenxid");
 	i_relpersistence = PQfnumber(res, "relpersistence");
 	i_isscannable = PQfnumber(res, "isscannable");
+	i_relpages = PQfnumber(res, "relpages");
 	i_owning_tab = PQfnumber(res, "owning_tab");
 	i_owning_col = PQfnumber(res, "owning_col");
 	i_reltablespace = PQfnumber(res, "reltablespace");
@@ -4557,6 +4674,7 @@ getTables(Archive *fout, int *numTables)
 		tblinfo[i].hastriggers = (strcmp(PQgetvalue(res, i, i_relhastriggers), "t") == 0);
 		tblinfo[i].hasoids = (strcmp(PQgetvalue(res, i, i_relhasoids), "t") == 0);
 		tblinfo[i].isscannable = (strcmp(PQgetvalue(res, i, i_isscannable), "t") == 0);
+		tblinfo[i].relpages = atoi(PQgetvalue(res, i, i_relpages));
 		tblinfo[i].frozenxid = atooid(PQgetvalue(res, i, i_relfrozenxid));
 		tblinfo[i].toast_oid = atooid(PQgetvalue(res, i, i_toastoid));
 		tblinfo[i].toast_frozenxid = atooid(PQgetvalue(res, i, i_toastfrozenxid));
@@ -4606,7 +4724,7 @@ getTables(Archive *fout, int *numTables)
 			resetPQExpBuffer(query);
 			appendPQExpBuffer(query,
 							  "LOCK TABLE %s IN ACCESS SHARE MODE",
-							  fmtQualifiedId(fout,
+							  fmtQualifiedId(fout->remoteVersion,
 										tblinfo[i].dobj.namespace->dobj.name,
 											 tblinfo[i].dobj.name));
 			ExecuteSqlStatement(fout, query->data);
@@ -4745,7 +4863,8 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 				i_conoid,
 				i_condef,
 				i_tablespace,
-				i_options;
+				i_options,
+				i_relpages;
 	int			ntups;
 
 	for (i = 0; i < numTables; i++)
@@ -4790,6 +4909,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 					 "pg_catalog.pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, i.indisclustered, "
+							  "t.relpages, "
 							  "c.contype, c.conname, "
 							  "c.condeferrable, c.condeferred, "
 							  "c.tableoid AS contableoid, "
@@ -4815,6 +4935,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 					 "pg_catalog.pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, i.indisclustered, "
+							  "t.relpages, "
 							  "c.contype, c.conname, "
 							  "c.condeferrable, c.condeferred, "
 							  "c.tableoid AS contableoid, "
@@ -4843,6 +4964,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 					 "pg_catalog.pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, i.indisclustered, "
+							  "t.relpages, "
 							  "c.contype, c.conname, "
 							  "c.condeferrable, c.condeferred, "
 							  "c.tableoid AS contableoid, "
@@ -4871,6 +4993,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 					 "pg_catalog.pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, i.indisclustered, "
+							  "t.relpages, "
 							  "c.contype, c.conname, "
 							  "c.condeferrable, c.condeferred, "
 							  "c.tableoid AS contableoid, "
@@ -4899,6 +5022,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 							  "pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, false AS indisclustered, "
+							  "t.relpages, "
 							  "CASE WHEN i.indisprimary THEN 'p'::char "
 							  "ELSE '0'::char END AS contype, "
 							  "t.relname AS conname, "
@@ -4925,6 +5049,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 							  "pg_get_indexdef(i.indexrelid) AS indexdef, "
 							  "t.relnatts AS indnkeys, "
 							  "i.indkey, false AS indisclustered, "
+							  "t.relpages, "
 							  "CASE WHEN i.indisprimary THEN 'p'::char "
 							  "ELSE '0'::char END AS contype, "
 							  "t.relname AS conname, "
@@ -4953,6 +5078,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 		i_indnkeys = PQfnumber(res, "indnkeys");
 		i_indkey = PQfnumber(res, "indkey");
 		i_indisclustered = PQfnumber(res, "indisclustered");
+		i_relpages = PQfnumber(res, "relpages");
 		i_contype = PQfnumber(res, "contype");
 		i_conname = PQfnumber(res, "conname");
 		i_condeferrable = PQfnumber(res, "condeferrable");
@@ -4995,6 +5121,7 @@ getIndexes(Archive *fout, TableInfo tblinfo[], int numTables)
 			parseOidArray(PQgetvalue(res, j, i_indkey),
 						  indxinfo[j].indkeys, INDEX_MAX_KEYS);
 			indxinfo[j].indisclustered = (PQgetvalue(res, j, i_indisclustered)[0] == 't');
+			indxinfo[j].relpages = atoi(PQgetvalue(res, j, i_relpages));
 			contype = *(PQgetvalue(res, j, i_contype));
 
 			if (contype == 'p' || contype == 'u' || contype == 'x')
@@ -12641,7 +12768,7 @@ createViewAsClause(Archive *fout, TableInfo *tbinfo)
 					  tbinfo->dobj.name);
 
 	/* Strip off the trailing semicolon so that other things may follow. */
-	Assert(PQgetvalue(res, 0, 0)[len-1] == ';');
+	Assert(PQgetvalue(res, 0, 0)[len - 1] == ';');
 	appendBinaryPQExpBuffer(result, PQgetvalue(res, 0, 0), len - 1);
 
 	PQclear(res);
@@ -12793,9 +12920,10 @@ dumpTableSchema(Archive *fout, TableInfo *tbinfo)
 			for (j = 0; j < tbinfo->numatts; j++)
 			{
 				/*
-			 * Normally, dump if it's locally defined in this table, and not
+				 * Normally, dump if it's locally defined in this table, and
-			 * dropped.  But for binary upgrade, we'll dump all the columns,
+				 * not dropped.  But for binary upgrade, we'll dump all the
-			 * and then fix up the dropped and nonlocal cases below.
+				 * columns, and then fix up the dropped and nonlocal cases
+				 * below.
 				 */
 				if (shouldPrintColumn(tbinfo, j))
 				{
@@ -12806,8 +12934,8 @@ dumpTableSchema(Archive *fout, TableInfo *tbinfo)
 											 !tbinfo->attrdefs[j]->separate);
 
 					/*
-				 * Not Null constraint --- suppress if inherited, except in
+					 * Not Null constraint --- suppress if inherited, except
-				 * binary-upgrade case where that won't work.
+					 * in binary-upgrade case where that won't work.
 					 */
 					bool		has_notnull = (tbinfo->notnull[j] &&
 											   (!tbinfo->inhNotNull[j] ||
@@ -12833,9 +12961,10 @@ dumpTableSchema(Archive *fout, TableInfo *tbinfo)
 					if (tbinfo->attisdropped[j])
 					{
 						/*
-					 * ALTER TABLE DROP COLUMN clears pg_attribute.atttypid,
+						 * ALTER TABLE DROP COLUMN clears
-					 * so we will not have gotten a valid type name; insert
+						 * pg_attribute.atttypid, so we will not have gotten a
-					 * INTEGER as a stopgap.  We'll clean things up later.
+						 * valid type name; insert INTEGER as a stopgap. We'll
+						 * clean things up later.
 						 */
 						appendPQExpBuffer(q, " INTEGER /* dummy */");
 						/* Skip all the rest, too */
@@ -12912,8 +13041,8 @@ dumpTableSchema(Archive *fout, TableInfo *tbinfo)
 			else if (!(tbinfo->reloftype && !binary_upgrade))
 			{
 				/*
-			 * We must have a parenthesized attribute list, even though empty,
+				 * We must have a parenthesized attribute list, even though
-			 * when not using the OF TYPE syntax.
+				 * empty, when not using the OF TYPE syntax.
 				 */
 				appendPQExpBuffer(q, " (\n)");
 			}
@@ -13853,8 +13982,8 @@ dumpSequence(Archive *fout, TableInfo *tbinfo)
 
 	/*
 	 * If the sequence is owned by a table column, emit the ALTER for it as a
-	 * separate TOC entry immediately following the sequence's own entry.
+	 * separate TOC entry immediately following the sequence's own entry. It's
-	 * It's OK to do this rather than using full sorting logic, because the
+	 * OK to do this rather than using full sorting logic, because the
 	 * dependency that tells us it's owned will have forced the table to be
 	 * created first.  We can't just include the ALTER in the TOC entry
 	 * because it will fail if we haven't reassigned the sequence owner to
@@ -14859,9 +14988,9 @@ findDumpableDependencies(ArchiveHandle *AH, DumpableObject *dobj,
 		else
 		{
 			/*
-			 * Object will not be dumped, so recursively consider its deps.
+			 * Object will not be dumped, so recursively consider its deps. We
-			 * We rely on the assumption that sortDumpableObjects already
+			 * rely on the assumption that sortDumpableObjects already broke
-			 * broke any dependency loops, else we might recurse infinitely.
+			 * any dependency loops, else we might recurse infinitely.
 			 */
 			DumpableObject *otherdobj = findObjectByDumpId(depid);
 
@@ -14884,22 +15013,21 @@ findDumpableDependencies(ArchiveHandle *AH, DumpableObject *dobj,
  *
  * Whenever the selected schema is not pg_catalog, be careful to qualify
  * references to system catalogs and types in our emitted commands!
+ *
+ * This function is called only from selectSourceSchemaOnAH and
+ * selectSourceSchema.
  */
 static void
 selectSourceSchema(Archive *fout, const char *schemaName)
 {
-	static char *curSchemaName = NULL;
 	PQExpBuffer query;
 
+	/* This is checked by the callers already */
+	Assert(schemaName != NULL && *schemaName != '\0');
+
 	/* Not relevant if fetching from pre-7.3 DB */
 	if (fout->remoteVersion < 70300)
 		return;
-	/* Ignore null schema names */
-	if (schemaName == NULL || *schemaName == '\0')
-		return;
-	/* Optimize away repeated selection of same schema */
-	if (curSchemaName && strcmp(curSchemaName, schemaName) == 0)
-		return;
 
 	query = createPQExpBuffer();
 	appendPQExpBuffer(query, "SET search_path = %s",
@@ -14910,9 +15038,6 @@ selectSourceSchema(Archive *fout, const char *schemaName)
 	ExecuteSqlStatement(fout, query->data);
 
 	destroyPQExpBuffer(query);
-	if (curSchemaName)
-		free(curSchemaName);
-	curSchemaName = pg_strdup(schemaName);
 }
 
 /*
@@ -15049,34 +15174,6 @@ myFormatType(const char *typname, int32 typmod)
 	return result;
 }
 
-/*
- * fmtQualifiedId - convert a qualified name to the proper format for
- * the source database.
- *
- * Like fmtId, use the result before calling again.
- */
-static const char *
-fmtQualifiedId(Archive *fout, const char *schema, const char *id)
-{
-	static PQExpBuffer id_return = NULL;
-
-	if (id_return)				/* first time through? */
-		resetPQExpBuffer(id_return);
-	else
-		id_return = createPQExpBuffer();
-
-	/* Suppress schema name if fetching from pre-7.3 DB */
-	if (fout->remoteVersion >= 70300 && schema && *schema)
-	{
-		appendPQExpBuffer(id_return, "%s.",
-						  fmtId(schema));
-	}
-	appendPQExpBuffer(id_return, "%s",
-					  fmtId(id));
-
-	return id_return->data;
-}
-
 /*
  * Return a column list clause for the given relation.
  *
@@ -15084,37 +15181,31 @@ fmtQualifiedId(Archive *fout, const char *schema, const char *id)
  * "", not an invalid "()" column list.
  */
 static const char *
-fmtCopyColumnList(const TableInfo *ti)
+fmtCopyColumnList(const TableInfo *ti, PQExpBuffer buffer)
 {
-	static PQExpBuffer q = NULL;
 	int			numatts = ti->numatts;
 	char	  **attnames = ti->attnames;
 	bool	   *attisdropped = ti->attisdropped;
 	bool		needComma;
 	int			i;
 
-	if (q)						/* first time through? */
+	appendPQExpBuffer(buffer, "(");
-		resetPQExpBuffer(q);
-	else
-		q = createPQExpBuffer();
-
-	appendPQExpBuffer(q, "(");
 	needComma = false;
 	for (i = 0; i < numatts; i++)
 	{
 		if (attisdropped[i])
 			continue;
 		if (needComma)
-			appendPQExpBuffer(q, ", ");
+			appendPQExpBuffer(buffer, ", ");
-		appendPQExpBuffer(q, "%s", fmtId(attnames[i]));
+		appendPQExpBuffer(buffer, "%s", fmtId(attnames[i]));
 		needComma = true;
 	}
 
 	if (!needComma)
 		return "";				/* no undropped columns */
 
-	appendPQExpBuffer(q, ")");
+	appendPQExpBuffer(buffer, ")");
-	return q->data;
+	return buffer->data;
 }
 
 /*

src/bin/pg_dump/pg_dump.h

@@ -252,6 +252,7 @@ typedef struct _tableInfo
 	/* these two are set only if table is a sequence owned by a column: */
 	Oid			owning_tab;		/* OID of table owning sequence */
 	int			owning_col;		/* attr # of column owning sequence */
+	int			relpages;
 
 	bool		interesting;	/* true if need to collect more data */
 
@@ -315,6 +316,7 @@ typedef struct _indxInfo
 	bool		indisclustered;
 	/* if there is an associated constraint object, its dumpId: */
 	DumpId		indexconstraint;
+	int			relpages;		/* relpages of the underlying table */
 } IndxInfo;
 
 typedef struct _ruleInfo
@@ -532,6 +534,7 @@ extern void sortDumpableObjects(DumpableObject **objs, int numObjs,
 					DumpId preBoundaryId, DumpId postBoundaryId);
 extern void sortDumpableObjectsByTypeName(DumpableObject **objs, int numObjs);
 extern void sortDumpableObjectsByTypeOid(DumpableObject **objs, int numObjs);
+extern void sortDataAndIndexObjectsBySize(DumpableObject **objs, int numObjs);
 
 /*
  * version specific routines

src/bin/pg_dump/pg_dump_sort.c

@@ -143,6 +143,96 @@ static void repairDependencyLoop(DumpableObject **loop,
 static void describeDumpableObject(DumpableObject *obj,
 					   char *buf, int bufsize);
 
+static int	DOSizeCompare(const void *p1, const void *p2);
+
+static int
+findFirstEqualType(DumpableObjectType type, DumpableObject **objs, int numObjs)
+{
+	int			i;
+
+	for (i = 0; i < numObjs; i++)
+		if (objs[i]->objType == type)
+			return i;
+	return -1;
+}
+
+static int
+findFirstDifferentType(DumpableObjectType type, DumpableObject **objs, int numObjs, int start)
+{
+	int			i;
+
+	for (i = start; i < numObjs; i++)
+		if (objs[i]->objType != type)
+			return i;
+	return numObjs - 1;
+}
+
+/*
+ * When we do a parallel dump, we want to start with the largest items first.
+ *
+ * Say we have the objects in this order:
+ * ....DDDDD....III....
+ *
+ * with D = Table data, I = Index, . = other object
+ *
+ * This sorting function now takes each of the D or I blocks and sorts them
+ * according to their size.
+ */
+void
+sortDataAndIndexObjectsBySize(DumpableObject **objs, int numObjs)
+{
+	int			startIdx,
+				endIdx;
+	void	   *startPtr;
+
+	if (numObjs <= 1)
+		return;
+
+	startIdx = findFirstEqualType(DO_TABLE_DATA, objs, numObjs);
+	if (startIdx >= 0)
+	{
+		endIdx = findFirstDifferentType(DO_TABLE_DATA, objs, numObjs, startIdx);
+		startPtr = objs + startIdx;
+		qsort(startPtr, endIdx - startIdx, sizeof(DumpableObject *),
+			  DOSizeCompare);
+	}
+
+	startIdx = findFirstEqualType(DO_INDEX, objs, numObjs);
+	if (startIdx >= 0)
+	{
+		endIdx = findFirstDifferentType(DO_INDEX, objs, numObjs, startIdx);
+		startPtr = objs + startIdx;
+		qsort(startPtr, endIdx - startIdx, sizeof(DumpableObject *),
+			  DOSizeCompare);
+	}
+}
+
+static int
+DOSizeCompare(const void *p1, const void *p2)
+{
+	DumpableObject *obj1 = *(DumpableObject **) p1;
+	DumpableObject *obj2 = *(DumpableObject **) p2;
+	int			obj1_size = 0;
+	int			obj2_size = 0;
+
+	if (obj1->objType == DO_TABLE_DATA)
+		obj1_size = ((TableDataInfo *) obj1)->tdtable->relpages;
+	if (obj1->objType == DO_INDEX)
+		obj1_size = ((IndxInfo *) obj1)->relpages;
+
+	if (obj2->objType == DO_TABLE_DATA)
+		obj2_size = ((TableDataInfo *) obj2)->tdtable->relpages;
+	if (obj2->objType == DO_INDEX)
+		obj2_size = ((IndxInfo *) obj2)->relpages;
+
+	/* we want to see the biggest item go first */
+	if (obj1_size > obj2_size)
+		return -1;
+	if (obj2_size > obj1_size)
+		return 1;
+
+	return 0;
+}
 
 /*
  * Sort the given objects into a type/name-based ordering

src/bin/pg_dump/pg_dumpall.c

@@ -1857,8 +1857,8 @@ connectDatabase(const char *dbname, const char *connection_string,
 	}
 
 	/*
-	 * Ok, connected successfully. Remember the options used, in the form of
+	 * Ok, connected successfully. Remember the options used, in the form of a
-	 * a connection string.
+	 * connection string.
 	 */
 	connstr = constructConnStr(keywords, values);
 

src/bin/pg_dump/pg_restore.c

@@ -71,6 +71,7 @@ main(int argc, char **argv)
 	RestoreOptions *opts;
 	int			c;
 	int			exit_code;
+	int			numWorkers = 1;
 	Archive    *AH;
 	char	   *inputFileSpec;
 	static int	disable_triggers = 0;
@@ -182,7 +183,7 @@ main(int argc, char **argv)
 				break;
 
 			case 'j':			/* number of restore jobs */
-				opts->number_of_jobs = atoi(optarg);
+				numWorkers = atoi(optarg);
 				break;
 
 			case 'l':			/* Dump the TOC summary */
@@ -313,7 +314,7 @@ main(int argc, char **argv)
 	}
 
 	/* Can't do single-txn mode with multiple connections */
-	if (opts->single_txn && opts->number_of_jobs > 1)
+	if (opts->single_txn && numWorkers > 1)
 	{
 		fprintf(stderr, _("%s: cannot specify both --single-transaction and multiple jobs\n"),
 				progname);
@@ -372,6 +373,18 @@ main(int argc, char **argv)
 	if (opts->tocFile)
 		SortTocFromFile(AH, opts);
 
+	/* See comments in pg_dump.c */
+#ifdef WIN32
+	if (numWorkers > MAXIMUM_WAIT_OBJECTS)
+	{
+		fprintf(stderr, _("%s: maximum number of parallel jobs is %d\n"),
+				progname, MAXIMUM_WAIT_OBJECTS);
+		exit(1);
+	}
+#endif
+
+	AH->numWorkers = numWorkers;
+
 	if (opts->tocSummary)
 		PrintTOCSummary(AH, opts);
 	else

src/tools/msvc/Mkvcbuild.pm

@@ -395,6 +395,7 @@ sub mkvcbuild
 	$psql->AddIncludeDir('src\bin\pg_dump');
 	$psql->AddIncludeDir('src\backend');
 	$psql->AddFile('src\bin\psql\psqlscan.l');
+	$psql->AddLibrary('ws2_32.lib');
 
 	my $pgdump = AddSimpleFrontend('pg_dump', 1);
 	$pgdump->AddIncludeDir('src\backend');
@@ -403,6 +404,7 @@ sub mkvcbuild
 	$pgdump->AddFile('src\bin\pg_dump\pg_dump_sort.c');
 	$pgdump->AddFile('src\bin\pg_dump\keywords.c');
 	$pgdump->AddFile('src\backend\parser\kwlookup.c');
+	$pgdump->AddLibrary('ws2_32.lib');
 
 	my $pgdumpall = AddSimpleFrontend('pg_dump', 1);
 
@@ -419,6 +421,7 @@ sub mkvcbuild
 	$pgdumpall->AddFile('src\bin\pg_dump\dumputils.c');
 	$pgdumpall->AddFile('src\bin\pg_dump\keywords.c');
 	$pgdumpall->AddFile('src\backend\parser\kwlookup.c');
+	$pgdumpall->AddLibrary('ws2_32.lib');
 
 	my $pgrestore = AddSimpleFrontend('pg_dump', 1);
 	$pgrestore->{name} = 'pg_restore';
@@ -426,6 +429,7 @@ sub mkvcbuild
 	$pgrestore->AddFile('src\bin\pg_dump\pg_restore.c');
 	$pgrestore->AddFile('src\bin\pg_dump\keywords.c');
 	$pgrestore->AddFile('src\backend\parser\kwlookup.c');
+	$pgrestore->AddLibrary('ws2_32.lib');
 
 	my $zic = $solution->AddProject('zic', 'exe', 'utils');
 	$zic->AddFiles('src\timezone', 'zic.c', 'ialloc.c', 'scheck.c',
@@ -572,6 +576,7 @@ sub mkvcbuild
 		$proj->AddIncludeDir('src\bin\psql');
 		$proj->AddReference($libpq, $libpgport, $libpgcommon);
 		$proj->AddResourceFile('src\bin\scripts', 'PostgreSQL Utility');
+		$proj->AddLibrary('ws2_32.lib');
 	}
 
 	# Regression DLL and EXE