Backup via rdump linux->solaris slow
My worksite has a backup system where a central solaris 2.5.1 server
rsh-es into various other UNIX boxes (Solaris, HPUX and AIX) and does
an rdump similar to this:
/sbin/rdump -0 -u -b 32 -s 1000000 -f ale:/dev/rmt/1hn /scm
I recently added a linux box to the network (debian 2.1, kernel
2.0.36), and the sysadmin tried adding two of it's ext2 partitions
(taking up most of a 2GB SCSI disk) to the backup routine. Rdump is
much slower on the linux box, to the point of being unusable. We found
storage speeds in the range of 80k/sec across a 10base-t connection. I
can ftp files between the two machines at ~900k/sec, and another
solaris box rdumps at ~690k/sec, so it's not the network. We've tried
playing with the blocksize figure, but we see only minor
changes. We've also seen errors like this in the backup report:
short read error from /dev/sda1: [block -2012730776]: count=1024, got=0
bread: lseek2 fails!
Any idea for a fix or at least a diagnosis? This is enough of a hiccup
that I will have to abandon linux as a platform for the task I have in
mind if I can't solve it. I'm not looking for suggestions of alternate
backup methods - for linux to work it has to fit within the current
backup paradigm.
Thanks in advance!
morgan
--
VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV
>> M o r g a n F l e t c h e r http://www.hahaha.org/morgan <<
>> Tibi gratias agimus quod nihil fumas. morgan@hahaha.org <<
Reply to: