Limiting i/o bandwith for pvmove?

Hi,

I have some problems with an overloaded disk-array. I want to move some volumes out of it, but the bandwidth pvmove normally takes is too heigh to remain stable.

Is there a way to limit the bandwidth usage of pvmove, like it is possible when reconstructing software raid mirris sets?

AVE!
Philipp Schmidt / phils...

Lbh unir whfg ivbyngrq gur Qvtvgny Zvyraavhz Pbclevtug Npg ol oernxvat gur
cebgrpgvba bs pbclevtugrq zngrevny.  Vs lbh ner abg n pvgvmra be erfvqrag bs
gur HFN, lbh evfx orvat vzcevfbarq naq uryq jvgubhg onvy sbe hc gb gjb jrrxf
hcba ragel gb gur HFN (c) Copyright 2001 by Hartmann Schaffer (signature only)

:wq

I wish there was! This problem annoys me too but nobody's been bothered enough about it to come up with a workaround yet...

(Maybe something could even be hacked together using ioband.)

Alasdair

[email protected]

I have had some success ensuring disk QoS using dm-ioband with XEN guests, but only on relatively ancient kernels (and versions of dm-ioband) so my results are inherently questionable.

I think ioband would work for this, the problem is its very difficult to construct test case scenarios like the one mentioned in this thread.

One would need machines of various architectures, chocked full of disks and programs to reproduce these types of scenarios. Ideally, you have half a dozen RAID cards to try, too.

This is a cause that I really hope to see companies who make storage solutions pick up, but none (to my knowledge) have.

Cheers,
--Tim

really ugly workaround: do it chunk-by-chunk...

for i in seq 0 10 10000; do

	pvmove overloaded01:$i-$[i+99] newarray:$i-$[i+99]
	sleep 5

done

or something like that...
you have to adjust offsets as needed.
you may want to restrict to a particular lv (-n yourlv). you can find out the PE mapping of your lvs with
pvs -o name,pvseg_start,pvseg_size,vg_name,lv_name
append /dev/pvname to restrict output on one pv

hth,

: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD� and LINBIT� are registered trademarks of LINBIT, Austria.