Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Neil Bothwick
On Thu, 30 Apr 2009 10:52:42 +0600, Mike Kazantsev wrote:

 So I thought there's gotta be something that fits these criterias, but
 so far I've found only dar and it seems quite slow and a bit unsuited
 for these needs.

What backup medium are you using? If hard disks, do you have a separate
machine for storing them? If so, BackupPC may suit your needs. The ebuild
in Portage is out of date but the latest version is available from
Bugzilla - http://bugs.gentoo.org/show_bug.cgi?id=141018

It allows restoration of individual files or directories, from any backup
point, and doesn't require any special software on the machines being
backed up, only SSH access. It can backup Windows and *nix machines.


-- 
Neil Bothwick

Waiter! There's a fly in my qagh!


signature.asc
Description: PGP signature


Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Mike Kazantsev
On Thu, 30 Apr 2009 08:25:38 +0100
Neil Bothwick n...@digimed.co.uk wrote:

 What backup medium are you using?

Oh my, I've managed to forget about it!


The medium is regular sata2 hard drives with ext3 filesystem on a
dedicated backup machine with quite rusty debian (etch) linux.
Most backed-up systems (that I care about) are actually freebsd 6, the
rest are linux. Most stored backups are 20-60 GB.

Main bottleneck here is the network - quite laggy 100 Mbps link,
because this backup server is quite far and isolated from the rest.
Also it's completely inaccessible from backed-up machines, aside from
reverse tunnels, which I rarely use as a dirty hacks.
And this link has tendency to go down every once in a while,
interrupting ongoing transfers.

That said, nightly backup should always be available, so the backups
are actually created on hotswap sata2 drives of each individual
machine and grabbed by backup server over ssh and, in some cases, nfs.

These days the scripts on the backup server quite frequently (10-50
times a day) connect to the other hosts and receive requests for
certain paths from stored backups. So they parse backups with tar
picking out given paths and pushes them back, as requested.
Needless to say, it is slow, hence the need.


Well, that's probably a bit more verbose than necessary, but the point
is that (I believe) the backups should be created right on backed-up
systems' hard disks, so:
  1. Have random access to backup storage.
  2. Prolonged io/cpu load is a bad thing.
  3. Compression (at least of gzip ratio) is a must, because of limited
storage on backed-up machines.
  4. In-backup seek times should be lower than tar (which is scanning
the whole file).
  5. I write py scripts for a living, so the question is really in a
backup format - transfer and storage structure is not the issue.

Easiest thing I've thought of is just to generate tar index on first
archive pass and then just skip to the recorded point in ungzipped
stream, feeding the rest to tar, stopping when necessary, but there's
no point to debug and maintain this system if there are better
solutions already.


 If hard disks, do you have a separate machine for storing them? If
 so, BackupPC may suit your needs. The ebuild in Portage is out of
 date but the latest version is available from Bugzilla -
 http://bugs.gentoo.org/show_bug.cgi?id=141018
 
 It allows restoration of individual files or directories, from any
 backup point, and doesn't require any special software on the
 machines being backed up, only SSH access. It can backup Windows and
 *nix machines.

Thanks, will check it out, but I'm afraid that live network backups
aren't the best solution in my case.


-- 
Mike Kazantsev // fraggod.net


signature.asc
Description: PGP signature


Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Joerg Schilling
Mike Kazantsev mike_kazant...@fraggod.net wrote:

 It seems that tar/gzip/bzip2 are almost universal solutions for
 unix-like system backups and we're using tar/gz combo to create backups
 from the dawn of times.
 But as the time goes by I stumble upon two misfits of such a combination
 more and more:

People on Linux who use the term tar are usually not talking about tar bug 
about gtar which is not 100% tar compatible and thus creates problems with
archive interchange.

 So I thought there's gotta be something that fits these criterias, but
 so far I've found only dar and it seems quite slow and a bit unsuited
 for these needs.

dar is using a nonstandard and proprietary archive format. 

Did you look at star?

Star is the oldest free tar implementation. It is 100% compatible to the 
standard and allows you to do incremental backups based on the POSIX.1-2001
archive format. Star includes support for all additional meta data.

Any at least POSIX.1-2001 compliant archiver is able to read the archives 
written by star and if you are ever need to restore a star based backup with
a different program, you only loose the ability to do incremental restores that 
deal with renamed/removed files.

ftp://ftp.berlios.de/pub/star/

The latest development source is in the schily source bundle at:

ftp://ftp.berlios.de/pub/schily/

Jörg

-- 
 EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
   j...@cs.tu-berlin.de(uni)  
   joerg.schill...@fokus.fraunhofer.de (work) Blog: 
http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily



Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Joerg Schilling
Neil Bothwick n...@digimed.co.uk wrote:

 BackupPC should cope with this. It uses rsync over SSH, so only needs to
 transfer new/changed files, and will restart where it left off if the
 connection fails (this happens to me sometimes when I switch off my
 laptop while it is backing up and the backup just restarts the next
 morning).

What rsync does can also be done by star by running something like:

umount /mnt
fssnap -F  ufs -d /export/nfs
rm /export/home/EXPORT-NFS.snap
sync
sleep 10
echo  /tmp/S.$$
svcadm disable svc:/network/nfs/server
mount -r `fssnap -F  ufs -o backing-store=/export/home/EXPORT-NFS.snap 
/export/nfs` /mnt
svcadm enable svc:/network/nfs/server

star bs=1m fs=256m -c -xdev -sparse -acl -link-dirs level=1 -cumulative 
dumpdate=/tmp/S.$$ fs-name=/export/nfs -
wtardumps tardumps=/etc/td-copy -C /mnt . | \
star bs=1m fs=256m -xpU -no-fsync -restore -time -C /export/nfs2


star supports enhanced file meta data, what does rsync?

Jörg

-- 
 EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
   j...@cs.tu-berlin.de(uni)  
   joerg.schill...@fokus.fraunhofer.de (work) Blog: 
http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily



Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Sebastian Günther
* Joerg Schilling (joerg.schill...@fokus.fraunhofer.de) [30.04.09 12:31]:
 Neil Bothwick n...@digimed.co.uk wrote:
 
  BackupPC should cope with this. It uses rsync over SSH, so only needs to
  transfer new/changed files, and will restart where it left off if the
  connection fails (this happens to me sometimes when I switch off my
  laptop while it is backing up and the backup just restarts the next
  morning).
 
 What rsync does can also be done by star by running something like:
 
 umount /mnt
 fssnap -F  ufs -d /export/nfs
 rm /export/home/EXPORT-NFS.snap
 sync
 sleep 10
 echo  /tmp/S.$$
 svcadm disable svc:/network/nfs/server
 mount -r `fssnap -F  ufs -o backing-store=/export/home/EXPORT-NFS.snap 
 /export/nfs` /mnt
 svcadm enable svc:/network/nfs/server
 
 star bs=1m fs=256m -c -xdev -sparse -acl -link-dirs level=1 -cumulative 
 dumpdate=/tmp/S.$$ fs-name=/export/nfs -
 wtardumps tardumps=/etc/td-copy -C /mnt . | \
 star bs=1m fs=256m -xpU -no-fsync -restore -time -C /export/nfs2
 
 

10 lines where 1 is sufficient? Not so userfriendly, and what is the 
benefit?

If space is not a problem the main benefit of rsync is that you have 
your backup simply on a filesystem, that you can mount anywhere you want 
it. And the plus is no programm needed to restore the data other than 
cp...

With rsnapshot it get even easier to configure your backup.

 star supports enhanced file meta data, what does rsync?
 
   -A, --acls  preserve ACLs (implies -p)
   -X, --xattrspreserve extended attributes

is all file meta data I have, so it is sufficient.

 Jörg
 

Sebastian

-- 
  Religion ist das Opium des Volkes.   |   _   ASCII ribbon campaign 
  Karl Marx  |  ( )   against HTML e-mail  
 s...@sti@N GÜNTHER   |   X   against M$ attachments
  mailto:sam...@guenther-roetgen.de  |  / \   www.asciiribbon.org  


pgpJCGuuTu1bG.pgp
Description: PGP signature


Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Neil Bothwick
On Thu, 30 Apr 2009 14:18:33 +0600, Mike Kazantsev wrote:

 The medium is regular sata2 hard drives with ext3 filesystem on a
 dedicated backup machine with quite rusty debian (etch) linux.
 Most backed-up systems (that I care about) are actually freebsd 6, the
 rest are linux. Most stored backups are 20-60 GB.
 
 Main bottleneck here is the network - quite laggy 100 Mbps link,
 because this backup server is quite far and isolated from the rest.
 Also it's completely inaccessible from backed-up machines, aside from
 reverse tunnels, which I rarely use as a dirty hacks.
 And this link has tendency to go down every once in a while,
 interrupting ongoing transfers.

BackupPC should cope with this. It uses rsync over SSH, so only needs to
transfer new/changed files, and will restart where it left off if the
connection fails (this happens to me sometimes when I switch off my
laptop while it is backing up and the backup just restarts the next
morning).

It is also space-efficient when backing up multiple machines, it uses
hard links to store only one copy of each file, no matter how many
machines have the same file. The lack of network access from the
clients to the server would mean you couldn't access the web interface
from the client you wished to restore to, but you could do that on the
backup server if necessary.


-- 
Neil Bothwick

The mechanic said I had blown a seal. I said, `Just fix the damn thing and
leave my private life out of it, OK?'


signature.asc
Description: PGP signature


Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Mike Kazantsev
On Thu, 30 Apr 2009 13:23:19 +0200
Sebastian Günther sam...@guenther-roetgen.de wrote:

 * Joerg Schilling (joerg.schill...@fokus.fraunhofer.de) [30.04.09 12:31]:
  What rsync does can also be done by star by running something like:
  
...
 
 10 lines where 1 is sufficient? Not so userfriendly, and what is the 
 benefit?

Most lines are used to freeze ufs as a snapshot, so they should be
there with rsync, as well.
Obvious benefit of a snapshot is consistency, and, luckily for me, no
users are involved in this process ;)


 If space is not a problem the main benefit of rsync is that you have 
 your backup simply on a filesystem, that you can mount anywhere you want 
 it. And the plus is no programm needed to restore the data other than 
 cp...

Sync is indeed a great idea, which should save tons of time and
resources, but some compression should still be necessary, even on
backup server, since most content should occupy 2x-10x space when
unpacked.

In fact, compressed write-enabled FS with snapshot capability plus rsync
is the closest thing to ideal backup as I can think of.
And in fact, it's called ZFS! ;)

Another sad fact is that the server is shared with a few other people,
so I can't just roll freebsd7 or solaris onto it, but I guess fuse-zfs
and compressed fuse filesystems should be worth a try indeed.

Besides, I've got some strange idea that to make squashfs you don't
really need its support in kernel...

All in all, more things to test out and think over, thanks.

-- 
Mike Kazantsev // fraggod.net


signature.asc
Description: PGP signature


Re: [gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-30 Thread Mike Kazantsev
On Thu, 30 Apr 2009 10:22:00 +0200
joerg.schill...@fokus.fraunhofer.de (Joerg Schilling) wrote:

 People on Linux who use the term tar are usually not talking about
 tar bug about gtar which is not 100% tar compatible and thus
 creates problems with archive interchange.

In fact, I'm more used to refer to freebsd tar as 'bsdtar', treating
GNU as standard ;)

 dar is using a nonstandard and proprietary archive format. 
 
 Did you look at star?

Actually that's the first thing I did, since it's closest (to tar)
implementation I know, but I haven't seen there the main reason why I've
decided to ditch tar - random access to files inside the archive (which
is stored on random-access media).
I don't have to pipe the archives sequentially, but if not for the
ever-increasing demand to read the contents, I'd have actually been
okay with tar.

Incremental backups are certainly handy feature to have, but I'm
relucant to use it, since it makes backups dependant on one another,
requiring additional logic for their storage.
When I think about that it seem like a great idea, as long as you
bundle them together all the time, so no increment gets lost, but then
my laziness and certain relucance to complicate things (so no one else
will be cursing me under his breath, sorting out why it lost some data)
always seem to get the upper hand :(

-- 
Mike Kazantsev // fraggod.net


signature.asc
Description: PGP signature


[gentoo-user] Random-access cross-platform FS_backup tool suggestions

2009-04-29 Thread Mike Kazantsev
Hello,

It seems that tar/gzip/bzip2 are almost universal solutions for
unix-like system backups and we're using tar/gz combo to create backups
from the dawn of times.
But as the time goes by I stumble upon two misfits of such a combination
more and more:

  1. It's quite inpractical to keep tens of tarballs for one backup.
  2. Seeking within single tens-hundreds-of-gigs tarball is suboptimal,
at least.
  3. Single thread operation.

At home, being a gentoo-only user (w/ gentoo-patched kernels), I've
solved the problem with squashfs - it keeps all the necessary
attributes, hardlinks, boasts multi-threaded creation and instant
access to any file within.

Alas, I can't use it on a production servers due to compatibility
issues - not a single linux here have support for it and changing
/ patching kernels is a bit of nuisance. Even worse, many systems that
need to be backed-up are FreeBSD.

So I thought there's gotta be something that fits these criterias, but
so far I've found only dar and it seems quite slow and a bit unsuited
for these needs.

Any suggestions?
Thanks.

-- 
Mike Kazantsev // fraggod.net


signature.asc
Description: PGP signature