Hi Everyone,

I have a production DB thats about 350 MB (single file) running on FB
(Version: LI-V6.2.908 Firebird 1.0) (SS 32 bit) on Redhat Linux 8.0 & I
am using a Athlon 2GHz server with 36 GB SCSI disk and 512 MB physical
RAM.

Something funny happens when I take a backup.

The queries experience a 300-700% performance drop immediately, even
after the backup process is complete. Now the really strange part. If I
delete the physical backup file so created query performance immediately
shoots up back to the earlier level.

I tried to take the backup from the Linux box directly by using the
command line gbak. The same results were observed.

eg. a simple query searching for a persons name takes 0 secs (some
milliseconds) and goes down to 6-7 seconds after backup.

I tied moving the backed up file to another location, but the same
problem persists. Only if I delete the physical backup file does the
performance come back. I monitored which files were being locked up
using lsof command but the backup file is not in use at all.

I have allocated nearly 10GB on the SCSI disk to the /tmp partition.

I am going crazy trying to figure this out. I am forced to monitor
backups on account of this and cannot automate the process. Earlier I
was on Win2K on the same machine and configuration and we have shifted
to RH8 about a month ago.

Can anyone help me please.

Warm regards,


RP