Have you ruled out the usual suspects - lack of memory and physically bad hard disk?
--- Supratim Sanyal, W1XMT 39.19151 N, 77.23432 W
First, hello to the list. and second on to why i am here.
Not sure if this is normal behavior or just my setup .
But i am getting constant disk thrashing for about a second
about every 3 seconds for 10 to 20 times after almost
any operation in any guest , be it linux winxp win2k bsd...
And with 5 not so quiet 2.5 inch 500 gig dell drives and workstation
close by , it gets pretty annoying, as i do not mind normal
access to disk but the periodic nature kind of bugs me.
Setup is devuan(debian without systemd) stable and tried testing with
same results.
Running ZFS in raidz 5 drives (tried both sync on and off) with a
dataset that contains
kvm raw and qcow2 images.
Tried every cache mode for virtual disks , thought it could be disk or
stat pooling
which is set to 3 sec by default, but changing period and turning all
pooling off does
not change the constant paging out to disk . Watching ntop and iotop
its qemu-system-x86-64
that is doing the write-(flushing), but why do 10 or 20 flushes after
a simple disk access
when one could suffice ? and that means its constantly doing this
disk thrashing
while using any guest , without even asking for any real disk access
like moving
files etc, any simple read to load something small in guest triggers this.
Went though all the config files for qemu and libvirtd , and hours in google
but cannot find anything relevant .
thanx in advance
dan
|