Add option to include WAL in base backup

When included, this makes the base backup a complete working
"clone" of the initial database, ready to have a postmaster
started against it without the need to set up any log archiving
or similar.

Magnus Hagander, reviewed by Fujii Masao and Heikki Linnakangas

Author: mhagander

doc/src/sgml/protocol.sgml

@@ -1460,7 +1460,7 @@ The commands accepted in walsender mode are:
   </varlistentry>
 
   <varlistentry>
-    <term>BASE_BACKUP [<literal>LABEL</literal> <replaceable>'label'</replaceable>] [<literal>PROGRESS</literal>] [<literal>FAST</literal>]</term>
+    <term>BASE_BACKUP [<literal>LABEL</literal> <replaceable>'label'</replaceable>] [<literal>PROGRESS</literal>] [<literal>FAST</literal>] [<literal>WAL</literal>]</term>
     <listitem>
      <para>
       Instructs the server to start streaming a base backup.
@@ -1505,6 +1505,18 @@ The commands accepted in walsender mode are:
          </para>
         </listitem>
        </varlistentry>
+
+       <varlistentry>
+        <term><literal>WAL</literal></term>
+        <listitem>
+         <para>
+          Include the necessary WAL segments in the backup. This will include
+          all the files between start and stop backup in the
+          <filename>pg_xlog</filename> directory of the base directory tar
+          file.
+         </para>
+        </listitem>
+       </varlistentry>
       </variablelist>
      </para>
      <para>
@@ -1561,7 +1573,10 @@ The commands accepted in walsender mode are:
        </listitem>
        <listitem>
         <para>
-         <filename>pg_xlog</> (including subdirectories)
+         <filename>pg_xlog</>, including subdirectories. If the backup is run
+         with wal files included, a synthesized version of pg_xlog will be
+         included, but it will only contain the files necessary for the
+         backup to work, not the rest of the contents.
         </para>
        </listitem>
       </itemizedlist>

doc/src/sgml/ref/pg_basebackup.sgml

@@ -144,6 +144,31 @@ PostgreSQL documentation
       </listitem>
      </varlistentry>
 
+     <varlistentry>
+      <term><option>-x</option></term>
+      <term><option>--xlog</option></term>
+      <listitem>
+       <para>
+        Includes the required transaction log files (WAL files) in the
+        backup. This will include all transaction logs generated during
+        the backup. If this option is specified, it is possible to start
+        a postmaster directly in the extracted directory without the need
+        to consult the log archive, thus making this a completely standalone
+        backup.
+       </para>
+       <note>
+        <para>
+         The transaction log files are collected at the end of the backup.
+         Therefore, it is necessary for the
+         <xref linkend="guc-wal-keep-segments"> parameter to be set high
+         enough that the log is not removed before the end of the backup.
+         If the log has been rotated when it's time to transfer it, the
+         backup will fail and be unusable.
+        </para>
+       </note>
+      </listitem>
+     </varlistentry>
+
      <varlistentry>
       <term><option>-Z <replaceable class="parameter">level</replaceable></option></term>
       <term><option>--compress=<replaceable class="parameter">level</replaceable></option></term>
@@ -164,7 +189,7 @@ PostgreSQL documentation
     <variablelist>
      <varlistentry>
       <term><option>-c <replaceable class="parameter">fast|spread</replaceable></option></term>
-      <term><option>--checkpoint <replaceable class="parameter">fast|spread</replaceable></option></term>
+      <term><option>--checkpoint=<replaceable class="parameter">fast|spread</replaceable></option></term>
       <listitem>
        <para>
         Sets checkpoint mode to fast or spread (default).
@@ -191,7 +216,10 @@ PostgreSQL documentation
         Enables progress reporting. Turning this on will deliver an approximate
         progress report during the backup. Since the database may change during
         the backup, this is only an approximation and may not end at exactly
-        <literal>100%</literal>.
+        <literal>100%</literal>. In particular, when WAL log is included in the
+        backup, the total amount of data cannot be estimated in advance, and
+        in this case the progress report will only count towards the total
+        amount of data without WAL.
        </para>
        <para>
         When this is enabled, the backup will start by enumerating the size of

src/backend/replication/basebackup.c

@@ -37,6 +37,7 @@ typedef struct
 	const char *label;
 	bool		progress;
 	bool		fastcheckpoint;
+	bool		includewal;
 }	basebackup_options;
 
 
@@ -46,11 +47,17 @@ static void _tarWriteHeader(char *filename, char *linktarget,
 				struct stat * statbuf);
 static void send_int8_string(StringInfoData *buf, int64 intval);
 static void SendBackupHeader(List *tablespaces);
-static void SendBackupDirectory(char *location, char *spcoid);
 static void base_backup_cleanup(int code, Datum arg);
 static void perform_base_backup(basebackup_options *opt, DIR *tblspcdir);
 static void parse_basebackup_options(List *options, basebackup_options *opt);
 
+/*
+ * Size of each block sent into the tar stream for larger files.
+ *
+ * XLogSegSize *MUST* be evenly dividable by this
+ */
+#define TAR_SEND_SIZE 32768
+
 typedef struct
 {
 	char	   *oid;
@@ -78,7 +85,10 @@ base_backup_cleanup(int code, Datum arg)
 static void
 perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
 {
-	do_pg_start_backup(opt->label, opt->fastcheckpoint);
+	XLogRecPtr	startptr;
+	XLogRecPtr	endptr;
+
+	startptr = do_pg_start_backup(opt->label, opt->fastcheckpoint);
 
 	PG_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum) 0);
 	{
@@ -87,12 +97,6 @@ perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
 		struct dirent *de;
 		tablespaceinfo *ti;
 
-
-		/* Add a node for the base directory */
-		ti = palloc0(sizeof(tablespaceinfo));
-		ti->size = opt->progress ? sendDir(".", 1, true) : -1;
-		tablespaces = lappend(tablespaces, ti);
-
 		/* Collect information about all tablespaces */
 		while ((de = ReadDir(tblspcdir, "pg_tblspc")) != NULL)
 		{
@@ -120,6 +124,10 @@ perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
 			tablespaces = lappend(tablespaces, ti);
 		}
 
+		/* Add a node for the base directory at the end */
+		ti = palloc0(sizeof(tablespaceinfo));
+		ti->size = opt->progress ? sendDir(".", 1, true) : -1;
+		tablespaces = lappend(tablespaces, ti);
 
 		/* Send tablespace header */
 		SendBackupHeader(tablespaces);
@@ -128,13 +136,102 @@ perform_base_backup(basebackup_options *opt, DIR *tblspcdir)
 		foreach(lc, tablespaces)
 		{
 			tablespaceinfo *ti = (tablespaceinfo *) lfirst(lc);
+			StringInfoData buf;
+
+			/* Send CopyOutResponse message */
+			pq_beginmessage(&buf, 'H');
+			pq_sendbyte(&buf, 0);		/* overall format */
+			pq_sendint(&buf, 0, 2);		/* natts */
+			pq_endmessage(&buf);
+
+			sendDir(ti->path == NULL ? "." : ti->path,
+					ti->path == NULL ? 1 : strlen(ti->path),
+					false);
 
-			SendBackupDirectory(ti->path, ti->oid);
+			/*
+			 * If we're including WAL, and this is the main data directory we
+			 * don't terminate the tar stream here. Instead, we will append
+			 * the xlog files below and terminate it then. This is safe since
+			 * the main data directory is always sent *last*.
+			 */
+			if (opt->includewal && ti->path == NULL)
+			{
+				Assert(lnext(lc) == NULL);
+			}
+			else
+				pq_putemptymessage('c');		/* CopyDone */
 		}
 	}
 	PG_END_ENSURE_ERROR_CLEANUP(base_backup_cleanup, (Datum) 0);
 
-	do_pg_stop_backup();
+	endptr = do_pg_stop_backup();
+
+	if (opt->includewal)
+	{
+		/*
+		 * We've left the last tar file "open", so we can now append the
+		 * required WAL files to it.
+		 */
+		uint32		logid,
+					logseg;
+		uint32		endlogid,
+					endlogseg;
+		struct stat statbuf;
+
+		MemSet(&statbuf, 0, sizeof(statbuf));
+		statbuf.st_mode = S_IRUSR | S_IWUSR;
+#ifndef WIN32
+		statbuf.st_uid = geteuid();
+		statbuf.st_gid = getegid();
+#endif
+		statbuf.st_size = XLogSegSize;
+		statbuf.st_mtime = time(NULL);
+
+		XLByteToSeg(startptr, logid, logseg);
+		XLByteToPrevSeg(endptr, endlogid, endlogseg);
+
+		while (true)
+		{
+			/* Send another xlog segment */
+			char		fn[MAXPGPATH];
+			int			i;
+
+			XLogFilePath(fn, ThisTimeLineID, logid, logseg);
+			_tarWriteHeader(fn, NULL, &statbuf);
+
+			/* Send the actual WAL file contents, block-by-block */
+			for (i = 0; i < XLogSegSize / TAR_SEND_SIZE; i++)
+			{
+				char		buf[TAR_SEND_SIZE];
+				XLogRecPtr	ptr;
+
+				ptr.xlogid = logid;
+				ptr.xrecoff = logseg * XLogSegSize + TAR_SEND_SIZE * i;
+
+				XLogRead(buf, ptr, TAR_SEND_SIZE);
+				if (pq_putmessage('d', buf, TAR_SEND_SIZE))
+					ereport(ERROR,
+							(errmsg("base backup could not send data, aborting backup")));
+			}
+
+			/*
+			 * Files are always fixed size, and always end on a 512 byte
+			 * boundary, so padding is never necessary.
+			 */
+
+
+			/* Advance to the next WAL file */
+			NextLogSeg(logid, logseg);
+
+			/* Have we reached our stop position yet? */
+			if (logid > endlogid ||
+				(logid == endlogid && logseg > endlogseg))
+				break;
+		}
+
+		/* Send CopyDone message for the last tar file */
+		pq_putemptymessage('c');
+	}
 }
 
 /*
@@ -147,6 +244,7 @@ parse_basebackup_options(List *options, basebackup_options *opt)
 	bool		o_label = false;
 	bool		o_progress = false;
 	bool		o_fast = false;
+	bool		o_wal = false;
 
 	MemSet(opt, 0, sizeof(*opt));
 	foreach(lopt, options)
@@ -180,6 +278,15 @@ parse_basebackup_options(List *options, basebackup_options *opt)
 			opt->fastcheckpoint = true;
 			o_fast = true;
 		}
+		else if (strcmp(defel->defname, "wal") == 0)
+		{
+			if (o_wal)
+				ereport(ERROR,
+						(errcode(ERRCODE_SYNTAX_ERROR),
+						 errmsg("duplicate option \"%s\"", defel->defname)));
+			opt->includewal = true;
+			o_wal = true;
+		}
 		else
 			elog(ERROR, "option \"%s\" not recognized",
 				 defel->defname);
@@ -316,26 +423,6 @@ SendBackupHeader(List *tablespaces)
 	pq_puttextmessage('C', "SELECT");
 }
 
-static void
-SendBackupDirectory(char *location, char *spcoid)
-{
-	StringInfoData buf;
-
-	/* Send CopyOutResponse message */
-	pq_beginmessage(&buf, 'H');
-	pq_sendbyte(&buf, 0);		/* overall format */
-	pq_sendint(&buf, 0, 2);		/* natts */
-	pq_endmessage(&buf);
-
-	/* tar up the data directory if NULL, otherwise the tablespace */
-	sendDir(location == NULL ? "." : location,
-			location == NULL ? 1 : strlen(location),
-			false);
-
-	/* Send CopyDone message */
-	pq_putemptymessage('c');
-}
-
 
 static int64
 sendDir(char *path, int basepathlen, bool sizeonly)
@@ -506,7 +593,7 @@ static void
 sendFile(char *filename, int basepathlen, struct stat * statbuf)
 {
 	FILE	   *fp;
-	char		buf[32768];
+	char		buf[TAR_SEND_SIZE];
 	size_t		cnt;
 	pgoff_t		len = 0;
 	size_t		pad;

src/backend/replication/repl_gram.y

@@ -71,6 +71,7 @@ Node *replication_parse_result;
 %token K_LABEL
 %token K_PROGRESS
 %token K_FAST
+%token K_WAL
 %token K_START_REPLICATION
 
 %type <node>	command
@@ -106,7 +107,7 @@ identify_system:
 			;
 
 /*
- * BASE_BACKUP [LABEL <label>] [PROGRESS] [FAST]
+ * BASE_BACKUP [LABEL '<label>'] [PROGRESS] [FAST] [WAL]
  */
 base_backup:
 			K_BASE_BACKUP base_backup_opt_list
@@ -136,7 +137,12 @@ base_backup_opt:
 				  $$ = makeDefElem("fast",
 						   (Node *)makeInteger(TRUE));
 				}
-
+			| K_WAL
+				{
+				  $$ = makeDefElem("wal",
+						   (Node *)makeInteger(TRUE));
+				}
+			;
 
 /*
  * START_REPLICATION %X/%X

src/backend/replication/repl_scanner.l

@@ -61,6 +61,7 @@ FAST			{ return K_FAST; }
 IDENTIFY_SYSTEM		{ return K_IDENTIFY_SYSTEM; }
 LABEL			{ return K_LABEL; }
 PROGRESS			{ return K_PROGRESS; }
+WAL			{ return K_WAL; }
 START_REPLICATION	{ return K_START_REPLICATION; }
 ","				{ return ','; }
 ";"				{ return ';'; }

src/backend/replication/walsender.c

@@ -105,7 +105,6 @@ static int	WalSndLoop(void);
 static void InitWalSnd(void);
 static void WalSndHandshake(void);
 static void WalSndKill(int code, Datum arg);
-static void XLogRead(char *buf, XLogRecPtr recptr, Size nbytes);
 static bool XLogSend(char *msgbuf, bool *caughtup);
 static void CheckClosedConnection(void);
 static void IdentifySystem(void);
@@ -649,8 +648,13 @@ WalSndKill(int code, Datum arg)
  *
  * XXX probably this should be improved to suck data directly from the
  * WAL buffers when possible.
+ *
+ * Will open, and keep open, one WAL segment stored in the global file
+ * descriptor sendFile. This means if XLogRead is used once, there will
+ * always be one descriptor left open until the process ends, but never
+ * more than one.
  */
-static void
+void
 XLogRead(char *buf, XLogRecPtr recptr, Size nbytes)
 {
 	XLogRecPtr	startRecPtr = recptr;