10000 Fix fsync-at-startup code to not treat errors as fatal. · luisacoding/postgres@d8179b0 · GitHub
[go: up one dir, main page]
Skip to content

Commit d8179b0

Browse files
tglsfdctglsfdc
committed
Fix fsync-at-startup code to not treat errors as fatal.
Commit 2ce439f introduced a rather serious regression, namely that if its scan of the data directory came across any un-fsync-able files, it would fail and thereby prevent database startup. Worse yet, symlinks to such files also caused the problem, which meant that crash restart was guaranteed to fail on certain common installations such as older Debian. After discussion, we agreed that (1) failure to start is worse than any consequence of not fsync'ing is likely to be, therefore treat all errors in this code as nonfatal; (2) we should not chase symlinks other than those that are expected to exist, namely pg_xlog/ and tablespace links under pg_tblspc/. The latter restriction avoids possibly fsync'ing a much larger part of the filesystem than intended, if the user has left random symlinks hanging about in the data directory. This commit takes care of that and also does some code beautification, mainly moving the relevant code into fd.c, which seems a much better place for it than xlog.c, and making sure that the conditional compilation for the pre_sync_fname pass has something to do with whether pg_flush_data works. I also relocated the call site in xlog.c down a few lines; it seems a bit silly to be doing this before ValidateXLOGDirectoryStructure(). The similar logic in initdb.c ought to be made to match this, but that change is noncritical and will be dealt with separately. Back-patch to all active branches, like the prior commit. Abhijit Menon-Sen and Tom Lane
1 parent d5442cb commit d8179b0

File tree

3 files changed

+256
-112
lines changed

3 files changed

+256
-112
lines changed

‎src/backend/access/transam/xlog.c

Lines changed: 12 additions & 42 deletions
Original file line numberDiff line numberDiff line change
@@ -866,8 +866,6 @@ static void WALInsertLockAcquireExclusive(void);
866866
static void WALInsertLockRelease(void);
867867
static void WALInsertLockUpdateInsertingAt(XLogRecPtr insertingAt);
868868

869-
static void fsync_pgdata(char *datadir);
870-
871869
/*
872870
* Insert an XLOG record represented by an already-constructed chain of data
873871
* chunks. This is a low-level routine; to construct the WAL record header
@@ -5951,18 +5949,6 @@ StartupXLOG(void)
59515949
(errmsg("database system was interrupted; last known up at %s",
59525950
str_time(ControlFile->time))));
59535951

5954-
/*
5955-
* If we previously crashed, there might be data which we had written,
5956-
* intending to fsync it, but which we had not actually fsync'd yet.
5957-
* Therefore, a power failure in the near future might cause earlier
5958-
* unflushed writes to be lost, even though more recent data written to
5959-
* disk from here on would be persisted. To avoid that, fsync the entire
5960-
* data directory.
5961-
*/
5962-
if (ControlFile->state != DB_SHUTDOWNED &&
5963-
ControlFile->state != DB_SHUTDOWNED_IN_RECOVERY)
5964-
fsync_pgdata(data_directory);
5965-
59665952
/* This is just to allow attaching to startup process with a debugger */
59675953
#ifdef XLOG_REPLAY_DELAY
59685954
if (ControlFile->state != DB_SHUTDOWNED)
@@ -5976,6 +5962,18 @@ StartupXLOG(void)
59765962
*/
59775963
ValidateXLOGDirectoryStructure();
59785964

5965+
/*
5966+
* If we previously crashed, there might be data which we had written,
5967+
* intending to fsync it, but which we had not actually fsync'd yet.
5968+
* Therefore, a power failure in the near future might cause earlier
5969+
* unflushed writes to be lost, even though more recent data written to
5970+
* disk from here on would be persisted. To avoid that, fsync the entire
5971+
* data directory.
5972+
*/
5973+ if (ControlFile->state != DB_SHUTDOWNED &&
5974+
ControlFile->state != DB_SHUTDOWNED_IN_RECOVERY)
5975+
SyncDataDirectory();
5976+
59795977
/*
59805978
* Initialize on the assumption we want to recover to the latest timeline
59815979
* that's active according to pg_control.
@@ -11602,31 +11600,3 @@ SetWalWriterSleeping(bool sleeping)
1160211600
XLogCtl->WalWriterSleeping = sleeping;
1160311601
SpinLockRelease(&XLogCtl->info_lck);
1160411602
}
11605-
11606-
/*
11607-
* Issue fsync recursively on PGDATA and all its contents.
11608-
*/
11609-
static void
11610-
fsync_pgdata(char *datadir)
11611-
{
11612-
if (!enableFsync)
11613-
return;
11614-
11615-
/*
11616-
* If possible, hint to the kernel that we're soon going to fsync the data
11617-
* directory and its contents.
11618-
*/
11619-
#if defined(HAVE_SYNC_FILE_RANGE) || \
11620-
(defined(USE_POSIX_FADVISE) && defined(POSIX_FADV_DONTNEED))
11621-
walkdir(datadir, pre_sync_fname);
11622-
#endif
11623-
11624-
/*
11625-
* Now we do the fsync()s in the same order.
11626-
*
11627-
* It's important to fsync the destination directory itself as individual
11628-
* file fsyncs don't guarantee that the directory entry for the file is
11629-
* synced.
11630-
*/
11631-
walkdir(datadir, fsync_fname);
11632-
}

0 commit comments

Comments
 (0)
0