You can manually change the I/O scheduler of the disks and if that works better for you, put a rule in udev.
Here is mine:
[root@engine rules.d]# cat /etc/udev/rules.d/90-default-io-scheduler.rules
ACTION=="add|change", KERNEL=="sd*[!0-9]", ATTR{queue/scheduler}="none"
ACTION=="add|change", KERNEL=="vd*[!0-9]", ATTR{queue/scheduler}="none"
[root@engine rules.d]# cat /sys/block/vda/queue/scheduler
[none] mq-deadline kyber
Best Regards,
Strahil Nikolov
Dear Users,I've successfully upgraded my 4 node hyperconverged system from 4.2.8 to 4.3.7.After upgrade everything seems to working fine, but the whole management system seems very slow.Spends many seconds when I clicking on "virtual machines" or I want to edit a virtual machines.The speed of vms and the IO is fine.It is running on a glusterfs (distributed replicate, on 3 node, 9 bricks). There are no errors, everything fine. but terrible slow:(The engine vm has 0.2-0.3 load.What can I do?Thanks in advance and I wish Happy New Year!Regards,Tibor