On Thu, 29 Apr 2004, David De Maeyer wrote:

>Hi all!
>
>We are facing a performance issue when taking a full backup of our mail
>server (>7000 users, 350 GB of data, running Debian linux).
>
>The backup server has no problem with the other linux clients. It is
>writing at around 100 KB/s... which is far from the 2 to 8 MB/s we are
>used to have on the other clients.
>
>It takes forever to backup all the user directories! And we simply
>can't find where the bottleneck is... just that there are a lot of I/O
>on the RAID on which these directories are stored.
>
>Would it help to increase/decrease the parallelism for the client
>and/or other clients (smaller filesystems)?

It would help a lot in responding to your message if you
explained what steps you have taken to investigate this
problem. Typically, a slow backup is the result of a mismatch
between the network interface card's setting and the setting
for the port on the switch to which the system is connected.
Make sure the network interface card is NOT set to auto-negotiate,
but is set to whatever the network port on the switch can handle.

Also, how is the file system(s) on that Linux box structured?
If you have that entire 350GB sitting on one mount point, you
can certainly improve the throughput by spreading the files
across several mount points, but this might entail a major
reconfiguration of your mail server configuration.

--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=