[Box Backup] Performance with Box Backup Server
Chris Wilson
boxbackup@boxbackup.org
Thu, 16 Apr 2009 23:02:38 +0100 (BST)
Hi Torsten,
On Thu, 16 Apr 2009, Torsten wrote:
> we have about 15 box backup clients connected to an dual core opteron
> with 1800 MHz. The box backup server runs inside an xen domU with 256
> megabytes ram assigned (Debian Etch).
>
> There is a heavy overload (permanently between 2 and 4) in the xen
> domain and also on the xen server dom0. This overload is not directly
> produced by box backup but by massive I/O activities.
>
> I have found this patch for box backup and it sounds good for me:
>
> http://boxbackup.org/trac/ticket/45
>
> Does anybody use this patch? Are there any problems with this?
I have not used it in production yet. I updated and tested it in
preparation for merging, but I discovered some unexplained test failures
that I did not have time to debug. It's possible that the patch simply
generates different, but still valid, diffs.
Alex Harper recently reported that he was seeing corruption on his store
server while using this patch:
http://lists.warhead.org.uk/pipermail/boxbackup/2009-January/004944.html
It's not clear to me whether or not the patch could have been responsible
for this.
> Any other ideas?
Many people have reported problems with I/O killing recent Linux kernels.
I've been able to reproduce it for myself; large disk writes (dd
if=/dev/zero of=bigfile bs=1M) basically bring the machine to a halt (load
average climbs over 150 and system is unresponsive and slow). My guess is
that reads are stalled waiting for the writes to complete, but more writes
are being added all the time so this never happens. You could try changing
the default I/O elevator on the kernel command line (changing at runtime
had no effect for me).
Cheers, Chris.
--
_____ __ _
\ __/ / ,__(_)_ | Chris Wilson <0000 at qwirx.com> - Cambs UK |
/ (_/ ,\/ _/ /_ \ | Security/C/C++/Java/Ruby/Perl/SQL Developer |
\__/_/_/_//_/___/ | We are GNU : free your mind & your software |