Remote 'dump' backup running extremely slowly

Ken Rossman rossman at columbia.edu
Mon Mar 1 14:14:01 UTC 2004


I have a pair of RedHat Linux 8.0 systems, which are set up as follows:

host1 owns a software RAIDed disk array, which has several NFS-exported 
volumes.
host2 is a utility host, which owns an SDLT tape drive.

I run a "dump" on host1 with "-f host2:/dev/nst0" as the target tape 
spec.
The backup runs fine (and fast) initially, and plows quickly through the
first several file systems (/, /boot, /usr).  Once it starts in on 
/export/home,
about midway through, it begins to slow down tremendously.

During this time, a "top" listing had been showing "dump" at the top of 
the CPU
usage listing, up until the backup began slowing down.  At that point, 
"kswapd"
had risen to the top of the list, and remains there for long periods of 
time,
at 99% CPU usage.  At this point, the backup appears to never finish 
(the data
rate decreases over time, and the target finish time increases).  I 
eventually
have to abort the backup.

Any idea what is going on here?

Thanks,
KR





More information about the redhat-list mailing list