10:11 p.m.
I agree, been checking some of my more disk intensive VMs this morning, switching them to
noop definitely improved responsiveness. All the virtio ones I’ve found were using
deadline (with RHEL/Centos guests), but some of the virt-scsi were using deadline and some
were noop, so I’m not sure of a definitive answer on that level yet.
For the hosts, it depends on what your backend is running. With a separate storage server
on my main cluster, it doesn’t matter what the hosts set for me. You mentioned you run
hyper converged, so I’d say it depends on what your disks are. If you’re using SSDs, go
none/noop as they don’t benefit from the queuing. If they are HDDs, I’d test cfq or
deadline and see which gave better latency and throughput to your vms. I’d guess you’ll
find deadline to offer better performance, but cfq to share better amongst multiple VMs.
Unless you use ZFS underneath, then go noop and let ZFS take care of it.
On Mar 18, 2019, at 2:05 PM, Strahil <hunter86_bg(a)yahoo.com>
wrote:
Hi Darrel,
Still, based on my experience we shouldn't queue our I/O in the VM, just to do the
same in the Host.
I'm still considering if I should keep deadline in my hosts or to switch to
'cfq'.
After all, I'm using Hyper-converged oVirt and this needs testing.
What I/O scheduler are you using on the host?
Best Regards,
Strahil Nikolov
On Mar 18, 2019 19:15, Darrell Budic <budic(a)onholyground.com> wrote:
Checked this on mine, see the same thing. Switching the engine to noop definitely feels
more responsive.
I checked on some VMs as well, it looks like virtio drives (vda, vdb….) get mq-deadline
by default, but virtscsi gets noop. I used to think the tuned profile for virtual-guest
would set noop, but apparently not…
-Darrell
On Mar 18, 2019, at 1:58 AM, Strahil Nikolov <hunter86_bg(a)yahoo.com
<mailto:hunter86_bg@yahoo.com>> wrote:
Hi All,
I have changed my I/O scheduler to none and here are the results so far:
Before (mq-deadline):
Adding a disk to VM (initial creation) START: 2019-03-17 16:34:46.709
Adding a disk to VM (initial creation) COMPLETED: 2019-03-17 16:45:17.996
After (none):
Adding a disk to VM (initial creation) START: 2019-03-18 08:52:02.xxx
Adding a disk to VM (initial creation) COMPLETED: 2019-03-18 08:52:20.xxx
Of course the results are inconclusive, as I have tested only once - but I feel the
engine more responsive.
Best Regards,
Strahil Nikolov
В неделя, 17 март 2019 г., 18:30:23 ч. Гринуич+2, Strahil <hunter86_bg(a)yahoo.com
<mailto:hunter86_bg@yahoo.com>> написа:
Dear All,
I have just noticed that my Hosted Engine has a strange I/O scheduler:
Last login: Sun Mar 17 18:14:26 2019 from 192.168.1.43 <http://192.168.1.43/>
[root@engine ~]# cat /sys/block/vda/queue/scheduler
[mq-deadline] kyber none
[root@engine ~]#
Based on my experience anything than noop/none is useless and performance degrading
for a VM.
Is there any reason that we have this scheduler ?
It is quite pointless to process (and delay) the I/O in the VM and then process (and
again delay) on Host Level .
If there is no reason to keep the deadline, I will open a bug about it.
Best Regards,
Strahil Nikolov
Dear All,
I have just noticed that my Hosted Engine has a strange I/O scheduler:
Last login: Sun Mar 17 18:14:26 2019 from 192.168.1.43 <http://192.168.1.43/>
[root@engine <mailto:root@engine> ~]# cat /sys/block/vda/queue/scheduler
[mq-deadline] kyber none
[root@engine <mailto:root@engine> ~]#
Based on my experience anything than noop/none is useless and performance degrading
for a VM.