pg_rewind: fsync target data directory.

Previously pg_rewind did not fsync any files. That's problematic, given
that the target directory is modified. If the database was started
afterwards, 2ce439f33 luckily already caused the data directory to be
synced to disk at postmaster startup; reducing the scope of the problem.

To fix, use initdb -S, at the end of the pg_rewind run. It doesn't seem
worthwhile to duplicate the code into pg_rewind, and initdb -S is
already used that way by pg_upgrade.

Reported-By: Andres Freund
Author: Michael Paquier, somewhat edited by me
Discussion: 20160310034352[email protected]
    CAB7nPqSytVG1o4S3S2pA1O=692ekurJ+fckW2PywEG3sNw54Ow@mail.gmail.com
Backpatch: 9.5, where pg_rewind was introduced

diff --git a/src/bin/pg_rewind/file_ops.c b/src/bin/pg_rewind/file_ops.c
index 193b72e9a2a396d7bd6246ad4c38c8d212dac50b..fceb5edd0273b2dc93991ac6ec84794aea6ef4b8 100644 (file)
--- a/src/bin/pg_rewind/file_ops.c
+++ b/src/bin/pg_rewind/file_ops.c
@@ -79,7 +79,6 @@ close_target_file(void)
                 dstpath, strerror(errno));
 
    dstfd = -1;
-   /* fsync? */
 }
 
 void

diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c
index 8e172f1618350d30a394a2dba8113d40ce69a634..9f2f08ebc4146cd1fcbb3b455f05df7a51e0446f 100644 (file)
--- a/src/bin/pg_rewind/pg_rewind.c
+++ b/src/bin/pg_rewind/pg_rewind.c
@@ -36,6 +36,7 @@ static void createBackupLabel(XLogRecPtr startpoint, TimeLineID starttli,
 static void digestControlFile(ControlFileData *ControlFile, char *source,
                  size_t size);
 static void updateControlFile(ControlFileData *ControlFile);
+static void syncTargetDirectory(const char *argv0);
 static void sanityChecks(void);
 static void findCommonAncestorTimeline(XLogRecPtr *recptr, TimeLineID *tli);
 
@@ -343,6 +344,9 @@ main(int argc, char **argv)
    ControlFile_new.state = DB_IN_ARCHIVE_RECOVERY;
    updateControlFile(&ControlFile_new);
 
+   pg_log(PG_PROGRESS, "syncing target data directory\n");
+   syncTargetDirectory(argv[0]);
+
    printf(_("Done!\n"));
 
    return 0;
@@ -572,3 +576,56 @@ updateControlFile(ControlFileData *ControlFile)
 
    close_target_file();
 }
+
+/*
+ * Sync target data directory to ensure that modifications are safely on disk.
+ *
+ * We do this once, for the whole data directory, for performance reasons.  At
+ * the end of pg_rewind's run, the kernel is likely to already have flushed
+ * most dirty buffers to disk. Additionally initdb -S uses a two-pass approach
+ * (only initiating writeback in the first pass), which often reduces the
+ * overall amount of IO noticeably.
+ */
+static void
+syncTargetDirectory(const char *argv0)
+{
+   int     ret;
+#define MAXCMDLEN (2 * MAXPGPATH)
+   char    exec_path[MAXPGPATH];
+   char    cmd[MAXCMDLEN];
+
+   /* locate initdb binary */
+   if ((ret = find_other_exec(argv0, "initdb",
+                              "initdb (PostgreSQL) " PG_VERSION "\n",
+                              exec_path)) < 0)
+   {
+       char        full_path[MAXPGPATH];
+
+       if (find_my_exec(argv0, full_path) < 0)
+           strlcpy(full_path, progname, sizeof(full_path));
+
+       if (ret == -1)
+           pg_fatal("The program \"initdb\" is needed by %s but was \n"
+                    "not found in the same directory as \"%s\".\n"
+                    "Check your installation.\n", progname, full_path);
+       else
+           pg_fatal("The program \"initdb\" was found by \"%s\"\n"
+                    "but was not the same version as %s.\n"
+                    "Check your installation.\n", full_path, progname);
+   }
+
+   /* only skip processing after ensuring presence of initdb */
+   if (dry_run)
+       return;
+
+   /* finally run initdb -S */
+   if (debug)
+       snprintf(cmd, MAXCMDLEN, "\"%s\" -D \"%s\" -S",
+                exec_path, datadir_target);
+   else
+       snprintf(cmd, MAXCMDLEN, "\"%s\" -D \"%s\" -S > \"%s\"",
+                exec_path, datadir_target, DEVNULL);
+
+   if (system(cmd) != 0)
+       pg_fatal("sync of target directory failed\n");
+}