Re: 5.1.21 Dell 2950 terrible swraid5 I/O performance with swraid on top of Perc 5/i raid0/jbod

[Marc MERLIN <marc@merlins.org>]

On Mon, Aug 19, 2019 at 11:18:13AM +0200, Paolo Valente wrote:
> Solving this kind of problem is one of the goals of the BFQ I/O scheduler [1].
> Have you tried?  If you want to, then start by swathing to BFQ in both the
> physical and the virtual block devices in your stack.
 
I sure was not aware of it, thank you for pointing it out.

> Thanks,
> Paolo
> 
> [1] https://algo.ing.unimo.it/people/paolo/BFQ/

I did the following below and when the swraid is rebuilding, I'm still
getting terrible overall throughput:
newmagic:~# hdparm -t /dev/md2
/dev/md2:
 HDIO_DRIVE_CMD(identify) failed: Inappropriate ioctl for device
  Timing buffered disk reads:   2 MB in  5.76 seconds = 355.42 kB/sec

I think things hang a bit less, which I suppose it good, but the system is
still unusable overall.

 
newmagic:~# modprobe bfq
newmagic:~# for i in /sys/block/*/queue/scheduler; do echo $i; echo bfq > $i; cat $i; done
/sys/block/bcache0/queue/scheduler
none
/sys/block/md0/queue/scheduler
none
/sys/block/md1/queue/scheduler
none
/sys/block/md2/queue/scheduler
none
/sys/block/md3/queue/scheduler
none                     
/sys/block/sda/queue/scheduler
[bfq] none
/sys/block/sdb/queue/scheduler
[bfq] none
/sys/block/sdc/queue/scheduler
[bfq] none
/sys/block/sdd/queue/scheduler
[bfq] none
/sys/block/sde/queue/scheduler
[bfq] none
/sys/block/sdf/queue/scheduler
[bfq] none
/sys/block/sdg/queue/scheduler
[bfq] none
/sys/block/sdh/queue/scheduler
[bfq] none
/sys/block/sdi/queue/scheduler
[bfq] none
/sys/block/sr0/queue/scheduler
[bfq] none


Thanks,
Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/