On 17/09/2024 08:58, B. Smith wrote:
Good evening,
I have a ZFS pool as a dedicated Bacula spool. The pool contains six 4TB
drives, configured as three mirrors of two striped disks. My tape drive
is LTO8. All the data is local to the server. When I despool without
simultaneously spooling another job, my despool rate is about 280
MB/sec. However, I noticed that when I allow a second job to spool at
the same time I despool, the despool rate goes down to as low as 75
MB/sec, with the drive stopping relatively frequently to refill the cache.
I'd like to try to figure out what the bottleneck is in this
configuration. I looked at using fio, but I don't want to test random
reads and writes... I want to test sequentially reading from one file on
the pool while sequentially writing to another. Can anyone suggest a
good way of testing this, or maybe point me to some other way of
determining the bottleneck? I realize it could be as simple as the disks
having poor simultaneous r/w performance, or it could be a limitation
with the backplane, HBA, or motherboard. Just trying to drill-down to
figure out what, if anything, I can upgrade.
Which ZFS are you using? Solaris, BSD, or Linux?
What's the HBA? NICs?
Are the disks SATA (so one-way only) or SAS (two-way)?
How much physical RAM does the system have?
I don't see this, but I am running Solaris with SAS disks in a RAIDZ
pool and 72Gb of RAM.
Cheers,
Gary B-)
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
