Ext3 vs Reiserfs benchmarks

Tested:

ext3 data=ordered
ext3 data=writeback
reiserfs
reiserfs notail

http://www.gurulabs.com/ext3-reiserfs.html

Any suggestions or comments appreciated.

Dax Kelson
Guru Labs

Did you try data=journal mode on ext3? For real-life workloads sync-IO workloads like mail (e.g. not benchmarks where the system is 100% busy) you can have considerable performance benefits from doing the sync IO directly to the journal instead of partly to the journal and partly to the rest of the filesystem.

The reason why "real life" is important here is because the data=journal mode writes all the files to disk twice - once to the journal and again to the filesystem, so you must have some "slack" in your disk bandwidth in order to benefit from this increased throughput on the part of the mail transport.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

I compared reiserfs with notails and with tails to ext3 in journaled mode about a month ago. Strangely enough the machine that was being built is eventually slated for a mail machine. I used postmark to simulate the mail environment.

Benchmarks are available here:
http://labs.zianet.com

Let me know if I am missing any info on there.

Steven

Yes, I saw this benchmark when it was first posted. It isn't clear from the web pages that you are using data=journal for ext3. Note that this is only a benefit for sync I/O workloads like mail and NFS, but not other types of usage. Also, for sync I/O workloads you can get a big boost by using an external journal device.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

postmark is an interesting workload, but it does not do fsync or renames on the working set, and postfix does lots of both while delivering. postmark does do a good job of showing the difference between lots of files in one directory (great for reiserfs) and lots of directories with fewer files in each (better for ext3).

Andreas Dilger already mentioned -o data=journal on ext3, you can try the beta reiserfs patches that add support for data=journal and data=ordered at:

ftp.suse.com/pub/people/mason/patches/data-logging

They improve reiserfs performance for just about everything, but data=journal is especially good for fsync/O_SYNC heavy workloads.

Andrew Morton sent me a benchmark of his that tries to simulate postfix. He has posted it to l-k before but a quick google search found dead links only, so I'm attaching it. What I like about his synctest is the results are consistent and you can play with various fsync/rename/unlink options.

-chris

/*
* Test and benchmark synchronous operations.
* stolen from Andrew Morton
*/

#undef _XOPEN_SOURCE /* MAP_ANONYMOUS */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <stdarg.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/resource.h>
#include <sys/wait.h>
#include <sys/mman.h>

/*
* Lots of yummy globals!
*/
char *progname, *dirname;
int verbose, use_fsync, use_osync;
int fsync_dir;
int n_threads = 1, n_iters = 100;
int *child_status;
int this_child_index;
int dir_fd;
int show_tids;
int threads_per_dir = 1;
int thread_group;
int do_unlink;
int rename_pass;

#define N_FILES 100
#define UNLINK_LAG 30
#define RENAME_PASSES 3

void show(char *fmt, ...)

{
	if (verbose) {
		va_list ap;

		va_start(ap, fmt);
		vfprintf(stdout, fmt, ap);
		fflush( stdout );
		va_end(ap);
	}
}

/*
* - Create a file.
* - Write some data to it
* - Maybe fsync() it.
* - Close it
* - Maybe fsync() its parent dir
* - rename() it.
* - maybe fsync() its parent dir
* - rename() it.
* - maybe fsync() its parent dir
* - rename() it.
* - maybe fsync() its parent dir
* - UNLINK_LAG files later, maybe unlink it.
* - maybe fsync() its parent dir
*
* Repeat the above N_FILES times
*/

char *mk_dirname(void)

{
	char *ret = malloc(strlen(dirname) + 64);

	sprintf(ret, "%s/%05d", dirname, thread_group);
	return ret;
}

char *mk_filename(int fileno)

{
	char *ret = malloc(strlen(dirname) + 64);

	sprintf(ret, "%s/%05d/%05d-%05d",
			dirname, thread_group, getpid(), fileno);
	return ret;
}

char *mk_new_filename(int fileno, int pass)

{
	char *ret = malloc(strlen(dirname) + 64);

	sprintf(ret, "%s/%05d/%02d-%05d-%05d",
			dirname, thread_group, pass, getpid(), fileno);
	return ret;
}

void sync_dir(void)

{
	if (fsync_dir) {
		show("fsync(%s)\n", dirname);
		if (fsync(dir_fd) < 0) {
			fprintf(stderr, "%s: failed to fsync dir `%s': %s\n",
				progname, dirname, strerror(errno));
			exit(1);
		}
	}
}

void make_dir(void)

{
	char *n = mk_dirname();

	show("mkdir(%s)\n", n);
	if (mkdir(n, 0777) < 0) {
		fprintf(stderr, "%s: Cannot make directory `%s': %s\n",
			progname, n, strerror(errno));
		exit(1);
	}
	free(n);
}

void remove_dir(void)

{
	char *n = mk_dirname();
	show("rmdir(%s)\n", n);
	rmdir(n);
	free(n);
}

void write_stuff_to(int fd, char *name)

{
	static char buf[500000];
	static int to_write = 5000;

	show("write %d bytes to `%s'\n", sizeof(buf), name);
	if (write(fd, buf, to_write) != to_write) {
		fprintf(stderr, "%s: failed to write %d bytes to `%s': %s\n",
			progname, to_write, name, strerror(errno));
		exit(1);
	}

	to_write *= 1.1;
	if (to_write > 250000)
		to_write = 5000;
}

void unlink_one_file(int fileno, int pass)

{
	if (do_unlink) {
		char *name = mk_new_filename(fileno, pass);

		show("unlink(%s)\n", name);
		if (unlink(name) < 0) {
			fprintf(stderr, "%s: failed to unlink `%s': %s\n",
				progname, name, strerror(errno));
			exit(1);
		}
		sync_dir();
		free(name);
	}
}

void do_one_file(int fileno)

{
	char *name = mk_filename(fileno);
	int fd, flags;

	flags = O_RDWR|O_CREAT|O_TRUNC;
	if (use_osync)
		flags |= O_SYNC;

	show("open(%s)\n", name);
	fd = open(name, flags, 0666);
	if (fd < 0) {
		fprintf(stderr, "%s: failed to create file `%s': %s\n",
			progname, name, strerror(errno));
		exit(1);
	}

	write_stuff_to(fd, name);

	if (use_fsync) {
		show("fsync(%s)\n", name);
		if (fsync(fd) < 0) {
			fprintf(stderr, "%s: failed to fsync `%s': %s\n",
				progname, name, strerror(errno));
			exit(1);
		}
	}

	show("close(%s)\n", name);
	if (close(fd) < 0) {
		fprintf(stderr, "%s: failed to close `%s': %s\n",
			progname, name, strerror(errno));
		exit(1);
	}

	sync_dir();

	for (rename_pass = 0; rename_pass < RENAME_PASSES; rename_pass++) {
		char *newname = mk_new_filename(fileno, rename_pass);

		show("rename(%s, %s)\n", name, newname);
		if (rename(name, newname) < 0) {
			fprintf(stderr,
				"%s: failed to rename `%s' to `%s': %s\n",
				progname, name, newname, strerror(errno));
			exit(1);
		}
		sync_dir();
		free(name);
		name = newname;
	}
	rename_pass--;
	free(name);
}

void do_child(void)

{
	int fileno;
	char *dn = mk_dirname();
	int dotcount;

	dir_fd = open(dn, O_RDONLY);
	if (dir_fd < 0) {
		fprintf(stderr, "%s: failed to open dir `%s': %s\n",
			progname, dn, strerror(errno));
		exit(1);
	}
	free(dn);

	dotcount = N_FILES / 10;
	if (dotcount == 0)
		dotcount = 1;

	for (fileno = 0; fileno < N_FILES; fileno++) {
		if (fileno % dotcount == 0) {
			printf(".");
			fflush(stdout);
		}
		do_one_file(fileno);
		if (fileno >= UNLINK_LAG)
			unlink_one_file(fileno - UNLINK_LAG, RENAME_PASSES - 1);
	}
	for (fileno = N_FILES - UNLINK_LAG; fileno < N_FILES; fileno++)
		unlink_one_file(fileno, RENAME_PASSES - 1);
}

void doit(void)

{
	int child;
	int children_left;

	child_status = (int *)mmap(	0,
				n_threads * sizeof(*child_status),
				PROT_READ|PROT_WRITE,
				MAP_SHARED|MAP_ANONYMOUS,
				-1,
				0);
	if (child_status == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	memset(child_status, 0, n_threads * sizeof(*child_status));

	thread_group = -1;
	for (this_child_index = 0;
			this_child_index < n_threads; this_child_index++)
	{
		if (this_child_index % threads_per_dir == 0) {
			thread_group++;
			make_dir();
		}

		if (fork() == 0) {
			int iter;

			for (iter = 0; iter < n_iters; iter++)
				do_child();
			child_status[this_child_index] = 1;
			exit(0);
		}
	}

	/* Parent */
	children_left = n_threads;
	while (children_left) {
		int status;

		if( wait3(&status, 0, 0) < 0 ) {
			if( errno != EINTR ) {
				perror("wait3");
				exit(1);
			}
			continue;
		}
		for (child = 0; child < n_threads; child++) {
			if (child_status[child] == 1) {
				child_status[child] = 2;
				printf("*");
				fflush(stdout);
				children_left--;
			}
		}
	}
	for (thread_group = 0; 
			thread_group < ( n_threads / threads_per_dir ); 
			thread_group++ )
		remove_dir();

	printf("\n");
}

void usage(void)

{
	fprintf(stderr,
		"Usage: %s [-fFosuv] [-p threads-pre-dir ][-n iters] [-t threads] dirname\n",
			progname);
	fprintf(stderr, "        -f:    Use fsync() on close\n"); 
	fprintf(stderr, "        -F:    Use fsync() on parent dir\n"); 
	fprintf(stderr, "        -n:    Number of iterations\n");
	fprintf(stderr, "        -o:    Open files O_SYNC\n");
	fprintf(stderr, "        -p:    Number of threads per directory\n");
	fprintf(stderr, "        -t:    Number of threads\n");
	fprintf(stderr, "        -u:    Unlink files during test\n");
	fprintf(stderr, "        -v:    Verbose\n"); 
	fprintf(stderr, "   dirname:    Directory to run tests in\n");
	exit(1);
}

int main(int argc, char *argv[])

{
	int c;

	progname = argv[0];
	while ((c = getopt(argc, argv, "vFfout:n:p:")) != -1) {
		switch (c) {
		case 'f':
			use_fsync++;
			break;
		case 'F':
			fsync_dir++;
			break;
		case 'n':
			n_iters = strtol(optarg, NULL, 10);
			break;
		case 'o':
			use_osync++;
			break;
		case 'p':
			threads_per_dir = strtol(optarg, NULL, 10);
			break;
		case 't':
			n_threads = strtol(optarg, NULL, 10);
			break;
		case 'u':
			do_unlink++;
			break;
		case 'v':
			verbose++;
			break;
		}
	}

	if (optind == argc)
		usage();
	dirname = argv[optind++];
	if (optind != argc)
		usage();

	doit();
	exit(0);
}

"it is clear that IF your server is stable and not prone to crashing, and/or you have the write cache on your hard drives battery backed, you should strongly consider using the writeback journaling mode of Ext3 versus ordered."

You probably want to suggest UPS there rather than battery backed disk cache, since the writeback caching is predominantly on the cpu side.

Daniel

Thanks for the feedback. Look for more testing from us soon addressing the suggestions brought up.

Dax

One more thing - can I just make the comment that testing freshly formatted filesystems is not going to show up ext2's real weaknesses, that happen to old filesystems - particularly those where the filesystem has been allowed to fill up.

I timed *15 minutes* for a system I admin to unlink a single 1G file on a fairly old ext2 filesystem the other day (perhaps ext3 would have improved this, I'm not sure). It took 30 minutes to scan a snort log directory log on ext2, but less than 2 minutes on reiser - and only 3 seconds once it was in the buffercache.

You are testing for a mail server - how many mailboxes are in your spool directory for the tests? Try it with about five to ten thousand mailboxes and see how your results vary. --
Sam Vilain, [email protected] WWW: http://sam.vilain.net/

    7D74 2A09 B2D3 C30F F78E      GPG: http://sam.vilain.net/sam.asc
    278A A425 30A9 05B5 2F13

Although Mr Chavez 'was democratically elected,' one had to bear in mind that 'Legitimacy is something that is conferred not just by a majority of the voters'"
- The office of George "Dubya" Bush commenting on the Venezuelan election

If your mail server can't get heirarchical mail spools right, get one that can.

Alan

Translation

"Yes, we know that there is no directory hashing in ext2/3. You'll have to find another solution to the problem, I'm afraid. Why not ease the burden on the filesystem by breaking up the task for it, and giving it to it in small pieces. That way it's much less likely to choke."

:-)

Sure, you could set up hierarchical mail spools. But it sure stinks of a temporary solution for a long-term problem. What about the next application that grows to massive proportions?

Hey, while I've got your attention, how do you go about debugging your kernel? I'm trying to add fair scheduling to the new O(1) scheduler, something of a token bucket filter counting jiffies used by a process/user/s_context (in scheduler_tick()) and tweaking their priority accordingly (in effective_prio()). It'd be really nice if I could run it under UML or something like that so I can trace through it with gdb, but I couldn't get the UML patch to apply to your tree. Any hints? --
Sam Vilain, [email protected] WWW: http://sam.vilain.net/

    7D74 2A09 B2D3 C30F F78E      GPG: http://sam.vilain.net/sam.asc
    278A A425 30A9 05B5 2F13

Actually there are several other reasons for it. It sucks a lot less when you need to use ls and friends to inspect part of the spool. It also makes it much easier to split the mail spool over multiple disks as it grows without having to backup/restore the spool area

JFS ?

The UML tree and my tree don't quite merge easily. Your best bet is to grab the Red Hat Limbo beta packages for the kernel source, which if I remember rightly are both -ac based and include the option to build UML

Alan

Another good reason is i_sem. If you've got more than one process doing something to that directory, you spend lots of time waiting for the semaphore. I think it was andrew that reminded me i_sem is held on fsync, so fync(dir) to make things safe after a rename can slow things down.

reiserfs only needs fsync(file), ext3 needs fsync(anything on the fs). If ext3 would promise to make fsync(file) sufficient forever, it might help the mta authors tune.

-chris

ext3 promises. This side-effect is bolted firmly into the design of ext3 and it's hard to see any way in which it will go away.

-

-aa ships with both uml and o1 scheduler. I need uml for everything non hardware related so expect it to be always uptodate there. However since I merged the O(1) scheduler there is the annoyance that sometime wakeup events don't arrive at least until kupdate reschedule or something like that (of course only with uml, not with real hardware). Also pressing keys is enough to unblock it. I didn't debugged it hard yet. Accoring to Jeff it's a problem with cli that masks signals.

Andrea

Amusingly, there IS directory hashing available for ext2 and ext3, and it is just as fast as reiserfs hashed directories. See:

http://people.nl.linux.org/~phillips/htree/paper/htree.html

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Faster, last time I checked. I really must test against XFS and JFS at some point.

Daniel

You learn something new every day. So, with that in mind - what has reiserfs got that ext2 doesn't?

- tail merging, giving much more efficient space usage for lots of small files.
- B*Tree allocation offering ``a 1/3rd reduction in internal fragmentation in return for slightly more complicated insertions and deletion algorithms'' (from the htree paper).
- online resizing in the main kernel (ext2 needs a patch -
http://ext2resize.sourceforge.net/).
- Resizing does not require the use of `ext2prepare' run on the filesystem while unmounted to resize over arbitrary boundaries.
- directory hashing in the main kernel

On the flipside, ext2 over reiserfs:

- support for attributes without a patch or 2.4.19-pre4+ kernel
- support for filesystem quotas without a patch
- there is a `dump' command (but it's useless, because it hangs when you run it on mounted filesystems - come on, who REALLY unmounts their filesystems for a nightly dump? You need a 3 way mirror to do it while guaranteeing filesystem availability...)

I'd be very interested in seeing postmark results without the hierarchical directory structure (which an unpatched postfix doesn't support), with about 5000 mailboxes with and without the htree patch (or with the htree patch but without that directory indexed, if that is possible). --
Sam Vilain, [email protected] WWW: http://sam.vilain.net/

    7D74 2A09 B2D3 C30F F78E      GPG: http://sam.vilain.net/sam.asc
    278A A425 30A9 05B5 2F13

Try to be the best of what you are, even if what you are is no good.
ASHLEIGH BRILLIANT

According to everybody, dump is deprecated (and it shouldn't work reliably with 2.4, in two words: "forget it")...

Mathieu Chouquet-Stringer              E-Mail : [email protected]
    It is exactly because a man cannot do a thing that he is a
                      proper judge of it.
                      -- Oscar Wilde

It's a shame, because `tar' doesn't save things like inode attributes and places unnecessary load on the VFS layer. It also takes considerably longer than dump did on one backup server I admin - like ~12 hours to back up ~26G in ~414k inodes to a tape capable of about 1MB/sec. But that's probably the old directory hashing thing again, there are some reeeeaaallllllly large directories on that machine...

Ah, the joys of legacy.
--
Sam Vilain, [email protected] WWW: http://sam.vilain.net/

    7D74 2A09 B2D3 C30F F78E      GPG: http://sam.vilain.net/sam.asc
    278A A425 30A9 05B5 2F13

If you think the United States has stood still, who built the
largest shopping center in the world?
RICHARD M NIXON

This needs to be "according to Linus, dump is deprecated". Given the interest Linus has manifested for backups, I wouldn't really rely on his statement :-)

Stelian.

Stelian Pop <[email protected]>
Alcove - http://www.alcove.com

Either way dump is not likely to give you a reliable backup when used with a 2.4.x kernel.

Gerhard

--
Gerhard Mack

[email protected]

<>< As a computer I find your faith in technology amusing.

Since you are so well informed, maybe you could share your knowledge with us.

I'm the dump maintainer, so I'll be very interested in knowing how comes that dump works for me and many other users... :-)

Stelian.

Stelian Pop <[email protected]>
Alcove - http://www.alcove.com

I'll save myself the trouble when Linus said it better than I could:

Note that dump simply won't work reliably at all even in 2.4.x: the buffer cache and the page cache (where all the actual data is) are not coherent. This is only going to get even worse in 2.5.x, when the directories are moved into the page cache as well.

So anybody who depends on "dump" getting backups right is already playing russian rulette with their backups. It's not at all guaranteed to get the right results - you may end up having stale data in the buffer cache that ends up being "backed up".

In other words you have a backup system that works some of the time or even most of the time... brilliant!

Gerhard

--
Gerhard Mack

[email protected]

<>< As a computer I find your faith in technology amusing.

just to clarify, the above implicitly assumes the fs is mounted read-write while you're dumping it. if the fs is mounted readonly or if it's unmounted, there is no problem with dumping it. Also note that dump has the same problem with read-write mounted fs also in 2.2, and I guess in 2.0 too, it's nothing new of 2.4, it just gets more visible the more logical dirty caches we have.

Andrea

Dump is a backup system that works 100% of the time when used as it was designed to: on unmounted filesystems (or mounted R/O).

It is indeed brilliant to have it work, even most of the time, in conditions it wasn't designed for.

Stelian.

Stelian Pop <[email protected]>
Alcove - http://www.alcove.com

Practical question: how do I get a file system mounted R/O for backup with dump without putting that system into single-user mode? Particularly when running automated backups, this is an issue. I cannot kill all writers (syslog, Postfix, INN, CVS server, ...) on my production machines just for the sake of taking a backup.

You don't.

This is where you have a filesystem where syslog, xinetd, blogd, bloatd-config-d2, raffle-ticketd DO NOT LIVE.

People forget so easily the wonders of multiple partitions.

On 07/16, Matthias Andree said something like:

--
Shawn Leas
[email protected]

So, do you live around here often?
-- Stephen Wright

I'm sorry, but I don't understand how it's going to change anything. For sure, it makes your life easier because you don't have to shutdown all your programs that have files opened in R/W mode. But in the end, you will have to shutdown something to remount the partition in R/O mode and usually you don't want or can't afford to do that.

Mathieu Chouquet-Stringer              E-Mail : [email protected]
    It is exactly because a man cannot do a thing that he is a
                      proper judge of it.
                      -- Oscar Wilde

In this case, can you use a RAID mirror or something, then break it?

Also, there's the LVM snapshot at the block layer someone already mentioned, which when used with smaller partions is less overhead. (less FS delta)

This problem isn't that complex.

On 07/16, Mathieu Chouquet-Stringer said something like:

--
Shawn Leas
[email protected]

I bought my brother some gift-wrap for Christmas. I took it to the Gift Wrap department and told them to wrap it, but in a different print so he would know when to stop unwrapping.
-- Stephen Wright

I agree but I guess that if Matthias asked the question that way, he probably meant he doesn't have a raid mirror or "something" (as you say)... If you didn't plan your install (meaning you don't have the nice raid or anything else), you're basically screwed...

Mathieu Chouquet-Stringer              E-Mail : [email protected]
    It is exactly because a man cannot do a thing that he is a
                      proper judge of it.
                      -- Oscar Wilde

All these "solutions" don't work out, I cannot remount R/O my partition, and LVM low-level snapshots or breaking a RAID mirror simply won't work out. I would have to remount r/o the partition to get a consistent image in the first place, so the first step must fail already...

Have you been reading my emails at all? LVM snapshots DO ensure that the snapshot filesystem is consistent for journaled filesystems.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Please apologize, I have been busy and only reading partial threads, and had not come across your LVM-snapshot related mails when I wrote the previous mail.

Matthias Andree

What kernel version is necessary to achieve this on production kernels (i. e. 2.4)?

Does "consistent" mean "fsck proof"?

Here's what I tried, on Linux-2.4.19-pre10-ac3 (IIRC) (ext3fs):

(from memory, history not available, different machine):
lvcreate --snapshot snap /dev/vg0/home
e2fsck -f /dev/vg0/snap
dump -0 ...

It reported zero dtime for one file and two bitmap differences.

Does "consistent" mean "consistent after you replay the log?" If so, that's still a losing game, because I cannot fsck the snapshot (it's R/O in the LVM case at least) to replay the journal -- and I don't assume dump 0.4b29 (which I'm using) goes fishing in the journal, but did not use the dump source code.

dump did not complain however, and given what e2fsck had to complain, I'd happily force mount such a file system when just a deletion has not completed.

Matthias Andree

That is because one critical piece is missing from 2.4, the VFS lock patch. It is part of the LVM sources at sistina.com. Chris Mason has been trying to get it in, but it is delayed until 2.4.19 is out.

You cannot mount a dirty ext3 filesystem from read-only media.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

On 07/19, Andreas Dilger said something like:

I thought you could "mount -t ext2" ext3 volumes, and thought you could force mount ext2.

I'm no Andreas Dilger, so don't take this like I'm disagreeing...

--
Shawn Leas
[email protected]

I went to the bank and asked to borrow a cup of money. They said, "What for?" I said, "I'm going to buy some sugar."
-- Stephen Wright

This is true if the ext3 filesystem is unmounted cleanly. Otherwise there is a flag in the superblock which tells the kernel it can't mount the filesystem because there is something there it doesn't understand (namely the dirty journal with all of the recent changes).

This flag (EXT3_FEATURE_INCOMPAT_RECOVERY) is cleared when the filesystem is unmounted properly, when e2fsck or a r/w mount recovers the journal, and not coincidentally when an LVM snapshot is created.

In case you are more curious, there are a couple of paragraphs in linux/Documentation/filesystems/ext2.txt about the compat flags, which are really one of the great features of ext2. You may think that an overstatement, but without the feature flags, none of the other enhancements that have been added to ext2 over the last few years (and in the next few years too) would have been so easily done.

As for mounting a dirty ext2 filesystem, yes that is possible with only a warning at mount time. That is why nobody has put much effort into adding the snapshot hooks into ext2 yet.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Well, there was a tail merging patch for ext2, but it has been dropped for now. In reality, any benchmarks with reiserfs (except the very-small-files case) will run with tail packing disabled because it kills performance.

Yes, I wrote it...

That is coming this summer. It will be part of some changes to support "meta blockgroups", and the resizing comes for free at the same time.

Probably will happen in 2.5, as Andrew is already testing htree support for ext3. It is also in the ext3 CVS tree for 2.4, so I wouldn't be surprised if it shows up in 2.4 also.

Well, the dump can only be inconsistent for files that are being changed during the dump itself. As for hanging the system, that would be a bug regardless of whether it was dump or "dd" reading from the block device. A bug related to this was fixed, probably in 2.4.19-preX somewhere.

Let me know what you find. It is possible to use an htree-patched kernel and not have indexed directories - just don't mount with "-o index". Note that there is a data-corrupting bug somewhere in the ext3 htree code, so I wouldn't suggest using indexed directories except for test.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

In article <[email protected]>,
Andreas Dilger <[email protected]> wrote:

| Well, the dump can only be inconsistent for files that are being changed
| during the dump itself.  As for hanging the system, that would be a bug
| regardless of whether it was dump or "dd" reading from the block device.
| A bug related to this was fixed, probably in 2.4.19-preX somewhere.

Any dump on a live f/s would seem to have the problem that files are changing as they are read and may not be consistant. I suppose there could be some kind of "fsync and journal lock" on a file, allowing all writes to a file to be journaled while the file is backed up. However, such things don't scale well for big files with lots of writes, and the file, while unchanging, may not be valid.

Backups of running files are best done by the application, like Oracle as a for-instance. Neither the o/s nor the backup can be sure when/if the data is in a valid state.

Tar has this problem, although not the same issues with data on the fly in buffers.

bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Hi Folks,

As an old dump user (dumpster?) I have to admit that we've avoided ext3 and Reiserfs because of this issue. We couldn't live without the "Tower of Hanoi".

I remember using, many years ago (SunOS 3.4), a patched dump binary that allowed safe dumps from live UFS filesystems. I don't remember all the details (it was 16-years ago) but this dump would compare somehow, files before and after writing to tape. If there was a difference it would back out the dumped file and preserve the consistency of the tape. I don't remember if it would go back and try the file again.

I haven't the foggest notion if this would work in these modern times, I'm just offering it as food for thought.

Regards,
Lew Wolfgang

[...]

dump(8) doesn't hang when dumping mounted filesystems. You are refering to a genuine bug which was fixed some time ago.

However, in some rare occasions, dump can save corrupted data when saving a mounted and generaly highly active filesystem. Even then, in 99% of the cases it doesn't really matter because the corrupted files will get saved by the next incremental dump.

Come on, who REALLY expects to have consistent backups without either unmounting the filesystem or using some snapshot techniques ?

Stelian.

Stelian Pop <[email protected]>
Alcove - http://www.alcove.com

The who uses [s|g]tar, cpio, afio, dsmc (Tivoli distributed storage manager), ...

Low-level snapshots don't do any good, they just freeze the "halfway there" on-disk structure.

But [s|g]tar, cpio, afio (don't know about dsmc) also freeze the "halfway there" data, but at the file level instead (application instead of filesystem)...

Stelian.

Stelian Pop <[email protected]>
Alcove - http://www.alcove.com

Not if some day somebody implements file system level snapshots for Linux. Until then, better have garbled file contents constrained to a file than random data as on-disk layout changes with hefty directory updates.

dsmc fstat()s the file it is currently reading regularly and retries the dump as the changes, and gives up if it is updated too often. Not sure about the server side, and certainly not a useful option for sequential devices that you directly write on. Looks like a cache for the biggest file is necessary.

or the blockdevice-level snapshots already implemented in Linux..

That would require three atomic steps:

1. mount read-only, flushing all pending updates
2. take snapshot
3. mount read-write

and then backup the snapshot. A snapshots of a live file system won't do, it can be as inconsistent as it desires -- if your corrupt target is moving or not, dumping it is not of much use.

Matthias Andree

Luckily, there is already an interface which does this - sync_supers_lockfs(), which the LVM code will use if it is patched in.

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Hi,

Well, couldn't we just kindof lock the file system so that while backing up no writes get through to the real filesystem? This will possibly require a lot of memory (or another space to write to), but it might be done?

Regards, Thunder

(Use http://www.ebb.org/ungeek if you can't decode)
------BEGIN GEEK CODE BLOCK------
Version: 3.12
GCS/E/G/S/AT d- s++:-- a? C++$ ULAVHI++++$ P++$ L++++(+++++)$ E W-$
N--- o?  K? w-- O- M V$ PS+ PE- Y- PGP+ t+ 5+ X+ R- !tv b++ DI? !D G
e++++ h* r--- y- 
------END GEEK CODE BLOCK------

This is all done already for both LVM and EVMS snapshots. The filesystem (ext3, reiserfs, XFS, JFS) flushes the outstanding operations and is frozen, the snapshot is created, and the filesystem becomes active again. It takes a second or less. Then dump will guarantee 100% correct backups of the snapshot filesystem. You would have to do a backup on the snapshot to guarantee 100% correctness even with tar.

Most people don't care, because they don't even do backups in the first place, until they have lost a lot of their data and they learn. Even without snapshots, while dump isn't guaranteed to be 100% correct for rapidly changing filesystems, I have been using it for years on both 2.2 and 2.4 without any problems on my home systems. I have even restored data from those same backups...

Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/

Hi,

Anyway, we could do that in parallel if we did it like that:

sync -> significant data is being written
lock -> data writes stay cached, but aren't written
snapshot
unlock -> data is getting written
now unmount the snapshout (clean it)
write the modified snapshot to disk...

Regards, Thunder

(Use http://www.ebb.org/ungeek if you can't decode)
------BEGIN GEEK CODE BLOCK------
Version: 3.12
GCS/E/G/S/AT d- s++:-- a? C++$ ULAVHI++++$ P++$ L++++(+++++)$ E W-$
N--- o?  K? w-- O- M V$ PS+ PE- Y- PGP+ t+ 5+ X+ R- !tv b++ DI? !D G
e++++ h* r--- y- 
------END GEEK CODE BLOCK------

Sure. On some machines, they will go with dsmc anyhow which reads the file and rereads if it changes under dsmc's hands.

Matthias Andree

I think I'm missing a part of this, the "a snapshot is created" sounds a lot like "here a miracle occurs." Where is this snapshot saved? And how do you take it in one sec regardless of f/s size? Is this one of those theoretical things which requires two mirrored copies of the f/s so you will still have RAID-1 after you break one? Or are changes journaled somewhere until the snapshot is transferred to external media? And how do you force applications to stop with their files in a valid state?

bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

LVM. Systems like LVM already provide a logical->physical block mapping on disk, so they might as well provide multiple mappings.

If the live filesystem writes to a particular disk block, the snapshot will keep referencing the old blocks while the filesystem gets to work on its own data. Copy on Write snapshots for block devices...

regards,

Rik

Bravely reimplemented by the knights who say "NIH".

http://www.surriel.com/		http://distro.conectiva.com/

It's really quite simple in theory to do proper backups. But you need to have application support to make it work in most cases. It would flow like this:

1. lock application(s), flush any outstanding transactions.
2. lock filesystems, flush any outstanding transactions.

3a. lock mirrored volume, flush any outstanding transactions, break

      mirror.
                --or--

3b. snapshot filesystem to another volume.

4. unlock volume

5. unlock filesystem

6. unlock application(s).

7. do backup against quiescent volume/filesystem.

In reality, people didn't lock filesystems (remount R/O) unless they had too (ClearCase, Oracle, any DBMS, etc are the exceptions), since the time hit was too much. The chances of getting a bad backup on user home directories or mail spools wasn't worth the extra cost to be sure to get a clean backup. For the exceptions, that's why god made backup windows and such. These days, those windows are miniscule, so the seven steps outlined above are what needs to happen these days for a trully reliable backup of important data.

John
John Stoffel - Senior Unix Systems Administrator - Lucent Technologies

	 [email protected] - http://www.lucent.com - 978-399-0479

Hi,

I do it like this:

-> Reconfigure port switch to use B server
-> Backup A server
-> Replay B server journals on A server
-> Switch to A server
-> Backup B server
-> Replay A server journals on B server
-> Reconfigure port switch to dynamic mode

Regards, Thunder

(Use http://www.ebb.org/ungeek if you can't decode)
------BEGIN GEEK CODE BLOCK------
Version: 3.12
GCS/E/G/S/AT d- s++:-- a? C++$ ULAVHI++++$ P++$ L++++(+++++)$ E W-$
N--- o?  K? w-- O- M V$ PS+ PE- Y- PGP+ t+ 5+ X+ R- !tv b++ DI? !D G
e++++ h* r--- y- 
------END GEEK CODE BLOCK------

Good summary. The problem is that 3a either requires a double morror or leaving the f/s un mirrored, and 3b can take a very long time for a big f/s.

In general mauch of this can be addressed by only backing up small f/s and using an application backup utility to backup the big stuff. Fortunately the most common problem apps are databases and and they include this capability.

bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

3b should be fairly quick since you only need to do an in-memory copy of some LVM metadata.

Rik

Bravely reimplemented by the knights who say "NIH".

http://www.surriel.com/		http://distro.conectiva.com/

Bill> Good summary. The problem is that 3a either requires a double Bill> morror or leaving the f/s un mirrored, and 3b can take a very Bill> long time for a big f/s.

Yup, 3a isn't a totally perfect solution, though triple mirrors (if you can afford them) work well. We actually do this for some servers where we can't afford the application down time of locking the DB for extended times, but we also don't have triple mirrors either. It's a tradeoff.

I really prefer 3b, since it's more efficient, faster, and more robust. To snapshot a filesystem, all you need to do is:

- create backing store for the snapshot, usually around 10-15% of the size of the original volume. Depends on volatility of data.
- lock the app(s).
- lock the filesystem and flush pending transactions.
- copy the metadata describing the filesystem
- insert a COW handler into the FS block write path
- mount the snapshot elsewhere
- unlock the FS
- unlock the app

Whenever the app writes a block into the FS, copy the original block to the backing store, then write the new block to storage.

All the backups see if the quiescent data store, so it can do a clean backup.

When you're done, just unmount the snapshot and delete it, then remove the backing store. There is an overhead for doing this, but it's better than having to unmirror/remirror whole block devices to do a backup. And cheaper in terms of disk space too.

Bill> In general mauch of this can be addressed by only backing up Bill> small f/s and using an application backup utility to backup the Bill> big stuff. Fortunately the most common problem apps are Bill> databases and and they include this capability.

Define what a small file system is these days, since it could be 100gb for some people. *grin*. It's a matter of making the tools scale well so that the data can be secured properly.

To do a proper backup requires that all layers talk to each other, and have some means of doing a RW lock and flush of pending transactions. If you have that, you can do it. If you don't, you need to either goto single user mode, re-mount RO, or pray.

John

Okay, other than the overhead and having enough filespace for Tbkup sec (min, hr, day) of operation this is practical. In general most times you would be doing an incremental, and the time would not be much.

Obviously a small f/s is one you can backup without operator intervantion to change media and in a reasonable time, which might be 10min..few hours depending on your taste. That's kind of my rule of thumb, you're welcome to suggest others, but if someone has to change media I can't call it small any more.

With some people, pray or ignore the problem are popular.

bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Or, to avoid the penalty of locking everything and bringing it down and stuff:

1. set a flag.

2. start backing up blocks (read them raw of course, don't want to load those stressed higher level systems)

3. If something wants to write to a block, quickly back up the old contents of the block before you write the new contents. Unless of course you've already backed up that block.

Of course, step 3 does place a bit more unschedulable load on the disk. Heck, when the backups have just started, you're doubling the latency of the devices. You can avoid this with a transaction journal; in fact, the cockier RDBMSes out there (eg, DMSII) don't even bother to do this and assume that your transaction journal is on a mirrored device - and hence there's no point in backing up the old data, you just want to do one sweep of the disk - and replay the journal to get current.

(note: implicit assumption: you're dealing with applications using synchronous I/O, where it needs to be written to all mirrors before it's trusted to be stored)

Ah, moot points - the Linux MD/LVM drivers are far too unsophisticated to have journal devices ;-)
--
Sam Vilain, [email protected] WWW: http://sam.vilain.net/

    7D74 2A09 B2D3 C30F F78E      GPG: http://sam.vilain.net/sam.asc
    278A A425 30A9 05B5 2F13

Law of Computability Applied to Social Sciences:
If at first you don't suceed, transform your data set.

In article <[email protected]>, Matthias Andree <[email protected]> wrote:

| dsmc fstat()s the file it is currently reading regularly and retries the
| dump as the changes, and gives up if it is updated too often. Not sure
| about the server side, and certainly not a useful option for sequential
| devices that you directly write on. Looks like a cache for the biggest
| file is necessary.

Which doesn't address the issue of data in files A, B and C, with indices in X and Y. This only works if you flush and freeze all the files at one time, making a perfect backup of one at a time results in corruption if the database is busy.

My favorite example is usenet news on INN, a bunch of circular spools, a linear history with two index files, 30-40k overview files, and all of it changing with perhaps 3.5MB/sec data and 20-50/sec index writes. Far better done with an application backup!

The point is, backups are hard, for many systems dump is optimal because it's fast. After that I like cpio (-Hcrc) but that's personal preference. All have fail cases on volatile data.

bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Right, but this would have to be taken up with Tivoli "do snapshot as dsmc starts, backup from snapshot and discard snapshot on exit"

In that case, when you are restoring from backups, you can also regenerate index files (at least with tradspool, I never looked at the "News in Dosen" aggregated spools like CNFS or whatever). It's really hard if you have .dir/.pag style dbm data bases that don't mirror some other single-file format.

Long ago (10-15 internet-years ago..) I followed testing of FFS-family of filesystems in Squid cache.

We noticed at Solaris machines using UFS, than when the directory data size grew above the number of blocks directly addressable by the direct-index pointers in the i-node, system speed plummeted. (Or perhaps it was something a bit smaller, like 32 kB)

Consider: 4 kB block size, 12 direct indexes: 48 kB directory size.

Spend 16 bytes for each file name + auxiliary data: 3000 files/subdirs

Optimal would be to store the files inside only the first block, e.g. the directory shall not grow over 4k (or 1k, or ..)

Name subdirs as: 00 thru 7F (128+2, 12 bytes ?) Possibly do that in 2 layers: 128^2 = 16384 subdirs, each with 50 long named users (even more files?): 820 000 users.

Tune the subdir hashing function to suit your application, and you should be happy.

Putting all your eggs in one basket (files in one directory) is not a smart thing.

/Matti Aarnio