Hello,
I have a lot of amanda backup files residing on 5 holding disks.
amflush can write to 2 tape drives in parallel.
My problem: amflush seems to pick backup files starting with the first
holding disk. When this is done amflush takes the backup files from the
second holding disk, and so on. Due to this the streams to the 2 drives
in most times come from the same holding disk.
The tapes can consume 300MB/s each whereas the holding disks (in this
concurrent-read-scenario) only provide 450MB/s in total. Hence the
drives do not get the data fast anough.
Is there a way to get amflush choose the next to-be-flushed backup file
by a different algorithm? Size, age, ... ? Even random would be better
than the current behaviour.
Thanks,
Sven
