Include the first valid listen address in pg_ctl to improve server start

"wait" detection and add postmaster start time to help determine if the
postmaster is actually using the specified data directory.

Author: bmomjian

doc/src/sgml/storage.sgml

@@ -117,8 +117,9 @@ last started with</entry>
 <row>
  <entry><filename>postmaster.pid</></entry>
  <entry>A lock file recording the current postmaster process id (PID),
- cluster data directory, port number, Unix domain socket directory,
- and shared memory segment ID</entry>
+ postmaster start time, cluster data directory, port number, user-specified
+ Unix domain socket directory, first valid listen_address host, and
+ shared memory segment ID</entry>
 </row>
 
 </tbody>

src/backend/port/ipc_test.c

@@ -104,7 +104,7 @@ on_exit_reset(void)
 }
 
 void
-RecordSharedMemoryInLockFile(unsigned long id1, unsigned long id2)
+AddToLockFile(int target_line, const char *str)
 {
 }
 

src/backend/port/sysv_shmem.c

@@ -198,9 +198,17 @@ InternalIpcMemoryCreate(IpcMemoryKey memKey, Size size)
 	/* Register on-exit routine to detach new segment before deleting */
 	on_shmem_exit(IpcMemoryDetach, PointerGetDatum(memAddress));
 
-	/* Record key and ID in lockfile for data directory. */
-	RecordSharedMemoryInLockFile((unsigned long) memKey,
-								 (unsigned long) shmid);
+	/*
+	 * Append record key and ID in lockfile for data directory. Format
+	 * to try to keep it the same length.
+	 */
+	{
+		char line[32];
+
+		sprintf(line, "%9lu %9lu\n", (unsigned long) memKey,
+									 (unsigned long) shmid);
+		AddToLockFile(LOCK_FILE_LINES, line);
+	}
 
 	return memAddress;
 }

src/backend/postmaster/postmaster.c

@@ -483,7 +483,8 @@ PostmasterMain(int argc, char *argv[])
 	int			status;
 	char	   *userDoption = NULL;
 	int			i;
-
+	bool		connection_line_output = false;
+	
 	MyProcPid = PostmasterPid = getpid();
 
 	MyStartTime = time(NULL);
@@ -860,10 +861,22 @@ PostmasterMain(int argc, char *argv[])
 										  UnixSocketDir,
 										  ListenSocket, MAXLISTEN);
 			else
+			{
 				status = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  UnixSocketDir,
 										  ListenSocket, MAXLISTEN);
+				/* must supply a valid listen_address for PQping() */
+				if (!connection_line_output)
+				{
+					char line[MAXPGPATH + 2];
+
+					sprintf(line, "%s\n", curhost);
+					AddToLockFile(LOCK_FILE_LINES - 1, line);
+					connection_line_output = true;
+				}
+			}
+				
 			if (status == STATUS_OK)
 				success++;
 			else
@@ -880,6 +893,10 @@ PostmasterMain(int argc, char *argv[])
 		pfree(rawstring);
 	}
 
+	/* Supply an empty listen_address line for PQping() */
+	if (!connection_line_output)
+		AddToLockFile(LOCK_FILE_LINES - 1, "\n");
+
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
 	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)

src/backend/utils/init/miscinit.c

@@ -46,7 +46,20 @@
 
 
 #define DIRECTORY_LOCK_FILE		"postmaster.pid"
-
+/*
+ *	The lock file contents are:
+ *
+ * line #
+ *		1	pid
+ *		2	postmaster start time
+ *		3	data directory
+ *		4	port #
+ *		5	user-specified socket directory
+ *			(the lines below are added later)
+ *		6	first valid listen_address
+ *		7	shared memory key
+ */
+	
 ProcessingMode Mode = InitProcessing;
 
 /* Note: we rely on this to initialize as zeroes */
@@ -678,7 +691,7 @@ CreateLockFile(const char *filename, bool amPostmaster,
 			   bool isDDLock, const char *refName)
 {
 	int			fd;
-	char		buffer[MAXPGPATH * 2 + 256];
+	char		buffer[MAXPGPATH * 3 + 256];
 	int			ntries;
 	int			len;
 	int			encoded_pid;
@@ -845,11 +858,10 @@ CreateLockFile(const char *filename, bool amPostmaster,
 		if (isDDLock)
 		{
 			char	   *ptr = buffer;
-			unsigned long id1,
-						id2;
+			unsigned long id1, id2;
 			int lineno;
 
-			for (lineno = 1; lineno <= 4; lineno++)
+			for (lineno = 1; lineno <= LOCK_FILE_LINES - 1; lineno++)
 			{
 				if ((ptr = strchr(ptr, '\n')) == NULL)
 				{
@@ -893,9 +905,10 @@ CreateLockFile(const char *filename, bool amPostmaster,
 	/*
 	 * Successfully created the file, now fill it.
 	 */
-	snprintf(buffer, sizeof(buffer), "%d\n%s\n%d\n%s\n",
+	snprintf(buffer, sizeof(buffer), "%d\n%ld\n%s\n%d\n%s\n",
 			 amPostmaster ? (int) my_pid : -((int) my_pid),
-			 DataDir, PostPortNumber, UnixSocketDir);
+			 (long) MyStartTime, DataDir, PostPortNumber,
+			 UnixSocketDir);
 	errno = 0;
 	if (write(fd, buffer, strlen(buffer)) != strlen(buffer))
 	{
@@ -1004,24 +1017,19 @@ TouchSocketLockFile(void)
 	}
 }
 
+
 /*
- * Append information about a shared memory segment to the data directory
- * lock file.
- *
- * This may be called multiple times in the life of a postmaster, if we
- * delete and recreate shmem due to backend crash.	Therefore, be prepared
- * to overwrite existing information.  (As of 7.1, a postmaster only creates
- * one shm seg at a time; but for the purposes here, if we did have more than
- * one then any one of them would do anyway.)
+ * Add lines to the data directory lock file.  This erases all following
+ * lines, but that is OK because lines are added in order.
  */
 void
-RecordSharedMemoryInLockFile(unsigned long id1, unsigned long id2)
+AddToLockFile(int target_line, const char *str)
 {
 	int			fd;
 	int			len;
 	int			lineno;
 	char	   *ptr;
-	char		buffer[MAXPGPATH * 2 + 256];
+	char		buffer[MAXPGPATH * 3 + 256];
 
 	fd = open(DIRECTORY_LOCK_FILE, O_RDWR | PG_BINARY, 0);
 	if (fd < 0)
@@ -1048,7 +1056,7 @@ RecordSharedMemoryInLockFile(unsigned long id1, unsigned long id2)
 	 * Skip over first four lines (PID, pgdata, portnum, socketdir).
 	 */
 	ptr = buffer;
-	for (lineno = 1; lineno <= 4; lineno++)
+	for (lineno = 1; lineno < target_line; lineno++)
 	{
 		if ((ptr = strchr(ptr, '\n')) == NULL)
 		{
@@ -1059,11 +1067,7 @@ RecordSharedMemoryInLockFile(unsigned long id1, unsigned long id2)
 		ptr++;
 	}
 
-	/*
-	 * Append key information.	Format to try to keep it the same length
-	 * always (trailing junk won't hurt, but might confuse humans).
-	 */
-	sprintf(ptr, "%9lu %9lu\n", id1, id2);
+	strlcat(buffer, str, sizeof(buffer));
 
 	/*
 	 * And rewrite the data.  Since we write in a single kernel call, this

src/bin/pg_ctl/pg_ctl.c

@@ -22,6 +22,7 @@
 
 #include <locale.h>
 #include <signal.h>
+#include <stdlib.h>
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <unistd.h>
@@ -138,6 +139,7 @@ static void read_post_opts(void);
 
 static PGPing test_postmaster_connection(bool);
 static bool postmaster_is_alive(pid_t pid);
+static time_t start_time;
 
 static char postopts_file[MAXPGPATH];
 static char pid_file[MAXPGPATH];
@@ -404,13 +406,13 @@ static PGPing
 test_postmaster_connection(bool do_checkpoint)
 {
 	int			portnum = 0;
-	char		socket_dir[MAXPGPATH];
+	char		host_str[MAXPGPATH];
 	char		connstr[MAXPGPATH + 256];
 	PGPing		ret = PQPING_OK;	/* assume success for wait == zero */
 	char	  **optlines;
 	int			i;
 
-	socket_dir[0] = '\0';
+	host_str[0] = '\0';
 	connstr[0] = '\0';
 
 	for (i = 0; i < wait_seconds; i++)
@@ -425,21 +427,25 @@ test_postmaster_connection(bool do_checkpoint)
 			 *		0	lock file created but status not written
 			 *		2	pre-9.1 server, shared memory not created
 			 *		3	pre-9.1 server, shared memory created
-			 *		4	9.1+ server, shared memory not created
-			 *		5	9.1+ server, shared memory created
+			 *		5	9.1+ server, listen host not created
+			 *		6	9.1+ server, shared memory not created
+			 *		7	9.1+ server, shared memory created
 			 *
 			 *	For pre-9.1 Unix servers, we grab the port number from the
 			 *	shmem key (first value on line 3).  Pre-9.1 Win32 has no
-			 *	written shmem key, so we fail.  9.1+ writes both the port
-			 *	number and socket address in the file for us to use.
+			 *	written shmem key, so we fail.  9.1+ writes connection
+			 *	information in the file for us to use.
 			 *	(PG_VERSION could also have told us the major version.)
 			 */
 
 			/* Try to read a completed postmaster.pid file */
 			if ((optlines = readfile(pid_file)) != NULL &&
 				optlines[0] != NULL &&
 				optlines[1] != NULL &&
-				optlines[2] != NULL)
+				optlines[2] != NULL &&
+				/* pre-9.1 server or listen_address line is present? */
+				(optlines[3] == NULL ||
+				 optlines[5] != NULL))
 			{				
 				/* A 3-line file? */
 				if (optlines[3] == NULL)
@@ -459,41 +465,63 @@ test_postmaster_connection(bool do_checkpoint)
 						return PQPING_NO_ATTEMPT;
 					}
 				}
-				else	/* 9.1+ server */
+				else
 				{
-					portnum = atoi(optlines[2]);
-	
-					/* Get socket directory, if specified. */
-					if (optlines[3][0] != '\n')
+					/*
+					 *	Easy check to see if we are looking at the right
+					 *	data directory:  Is the postmaster older than this
+					 *	execution of pg_ctl?  Subtract 2 seconds to account
+					 *	for possible clock skew between pg_ctl and the
+					 *	postmaster.
+					 */
+					if (atol(optlines[1]) < start_time - 2)
 					{
-						/*
-						 *	While unix_socket_directory can accept relative
-						 *	directories, libpq's host must have a leading slash
-						 *	to indicate a socket directory.
-						 */
-						if (optlines[3][0] != '/')
-						{
-							write_stderr(_("%s: -w option cannot use a relative socket directory specification\n"),
-										 progname);
-							return PQPING_NO_ATTEMPT;
-						}
-						strlcpy(socket_dir, optlines[3], MAXPGPATH);
-						/* remove newline */
-						if (strchr(socket_dir, '\n') != NULL)
-							*strchr(socket_dir, '\n') = '\0';
+						write_stderr(_("%s: this data directory is running an older postmaster\n"),
+									 progname);
+						return PQPING_NO_ATTEMPT;
 					}
-				}
+						
+					portnum = atoi(optlines[3]);
 
+					/*
+					 *	Determine the proper host string to use.
+					 */
+#ifdef HAVE_UNIX_SOCKETS
+					/*
+					 *	Use socket directory, if specified.  We assume if we
+					 *	have unix sockets, the server does too because we
+					 *	just started the postmaster.
+					 */
+					/*
+					 *	While unix_socket_directory can accept relative
+					 *	directories, libpq's host must have a leading slash
+					 *	to indicate a socket directory.
+					 */
+					if (optlines[4][0] != '\n' && optlines[4][0] != '/')
+					{
+						write_stderr(_("%s: -w option cannot use a relative socket directory specification\n"),
+									 progname);
+						return PQPING_NO_ATTEMPT;
+					}
+					strlcpy(host_str, optlines[4], sizeof(host_str));
+#else
+					strlcpy(host_str, optlines[5], sizeof(host_str));
+#endif
+					/* remove newline */
+					if (strchr(host_str, '\n') != NULL)
+						*strchr(host_str, '\n') = '\0';
+				}
+				
 				/*
 				 * We need to set connect_timeout otherwise on Windows the
 				 * Service Control Manager (SCM) will probably timeout first.
 				 */
 				snprintf(connstr, sizeof(connstr),
 						 "dbname=postgres port=%d connect_timeout=5", portnum);
 
-				if (socket_dir[0] != '\0')
+				if (host_str[0] != '\0')
 					snprintf(connstr + strlen(connstr), sizeof(connstr) - strlen(connstr),
-						" host='%s'", socket_dir);
+						" host='%s'", host_str);
 			}
 		}
 
@@ -1756,6 +1784,7 @@ main(int argc, char **argv)
 
 	progname = get_progname(argv[0]);
 	set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("pg_ctl"));
+	start_time = time(NULL);
 
 	/*
 	 * save argv[0] so do_start() can look for the postmaster if necessary. we

src/include/miscadmin.h

@@ -348,11 +348,11 @@ extern PGDLLIMPORT bool process_shared_preload_libraries_in_progress;
 extern char *shared_preload_libraries_string;
 extern char *local_preload_libraries_string;
 
+#define LOCK_FILE_LINES		7
 extern void CreateDataDirLockFile(bool amPostmaster);
 extern void CreateSocketLockFile(const char *socketfile, bool amPostmaster);
 extern void TouchSocketLockFile(void);
-extern void RecordSharedMemoryInLockFile(unsigned long id1,
-							 unsigned long id2);
+extern void AddToLockFile(int target_line, const char *str);
 extern void ValidatePgVersion(const char *path);
 extern void process_shared_preload_libraries(void);
 extern void process_local_preload_libraries(void);