Hello Unix Gurus:
We have a Unix cluster consisting of two AlphaServer 4200 systems running
DU 4.0d. This cluster serves as our central academic computer system for
us here at Temple University. Right now, there are over 24,000 accounts on
the system. Most people use our cluster to use e-mail either with Pine,
Elm, or a POPmail client, Web page hosting, and other Internet services.
We run a backup server with Networker 4.4 on an AlphaServer 8200 that
handles backups for about 15 different client systems, including our Unix
cluster. Our other client systems all have very light requirements for
backups, compared to our Unix cluster. Right now, the files on the Unix
cluster, including the mail/spool files are backed up incrementally each
night with a full backup done on the first day of each month. We have a
TL812 jukebox on our 8200 for our backups that holds 47 tapes and they are
increasingly getting filled faster and faster as computer usage increases
here.
Much to our chagrin, we lost all our mail/spool files on our Cluster on
Monday afternoon. We recovered most of the files back via NSR. Messages
that came in after our daily backup ran on Sunday at 11:00pm and the time
the mail/spool files were lost permanently. Recovering these files was a
real chore because new mail/spool files were being created for many people
(due to incoming mail) while the recovery was in progress. We had to
recover around 18,100 files for a total 17 GIG worth of data. This
recovery took at least three hours.
What I am wondering is how others handle backups of their /mail/spool
files on heavily active systems like our's. Are incrementals for just the
Cluster's mail/spool files typically run at 10-20 GIGs and take 3-4 hours
to complete. We are thinking of doing incrementals of just the /mail/spool
files twice a day rather than once. I am just wondering what this will do
in terms of consuming more disk space for the additional NSR index files
on our backup server. We give NSR an entire 18 GIG raided disk now and it
runs at around 70% capacity. I am also pretty good at keeping the index
files pruned.
Had we lost our files on our /home volume, it would have been a much
easier task to recover them, but the dynamic nature of the /mail/spool
entries is a real challenge to handle. We might also be in the position to
buy an additional tape library to handle our increased backup needs. In
any event, I am very curious to know what other people who work in this
type of environment do to backup their system's mail/spool files.
Thanks
Received on Wed Feb 24 1999 - 21:01:55 NZDT