Limit IO for each VM? or max disk MB/s available for a VM?

J

jmjosebest

Renowned Member
Jan 16, 2009
192
39
93
Hi, it's possible add any limit in MB/s that a qemu is available to write into the storage?
Thanks!
 
not yet, but in KVM there is such a functionality so I assume sooner or later this will find its way into Proxmox VE.
 
maybe a stupid question, but is it possible to start a vm with lower/higher ionice priority?

thanks
u.
 
what if you start a kvm process with lower ionice priority?
won't that work this way?
 
hopefully someone with bigger skills than mine will test that, and i think it can be eysily implemented to pve, if it works
 
You do not need to start a process with ionice to use ionice.

Grab the pid of your kvm process and use ionice to adjust things.

Highest Priority:
Code: 
 ionice -c 2 -n 0 -p pidhere

Lowest Priority:
Code: 
 ionice -c 2 -n 7 -p pidhere

Doing that you could at least see if this works as expected or respond to a problem.
 
i just wanted to test large file copy on two windows vms, first WITHOUT setting any io priority and the result was surprising:

while copying the same file on both vm the second (w7) took almost all the io bandwidth from the first (xp)

root@genya:~# cat /etc/pve/qemu-server/108.conf
acpi: 1
boot: dc
bootdisk: virtio0
cores: 2
cpuunits: 1000
freeze: 0
ide2: none,media=cdrom
keyboard: hu
kvm: 1
memory: 3072
name: masodik
net0: virtio=62:88:9A:D1:37:2A,bridge=vmbr1
onboot: 1
ostype: wxp
sockets: 2
vga: std
virtio0: virt:108/vm-108-disk-2.raw
virtio1: virt:108/vm-108-disk-1.raw

root@genya:~# cat /etc/pve/qemu-server/109.conf
args: -bios /usr/share/kvm/bios.bin.asus
boot: dc
bootdisk: virtio0
cores: 2
ide2: none,media=cdrom
keyboard: hu
memory: 3072
name: w7
net0: virtio=92:38:02:6E:DA:05,bridge=vmbr1
ostype: win7
sockets: 1
vga: std
virtio0: virt:109/vm-109-disk-1.raw
virtio1: virt:109/vm-109-disk-2.raw
 
I am currently testing with ionice ...
The best way to do it is to start the VM with ionice , something like : "ionice -c3 qm start <vmid>"

It has shown some promising results , and if combined with a nice it can do wonders for the cpu to . ( nice -n14 ionice -c2 -n7 qm start <vmid> )
It is a bit of a problem to auto start .. but a simple bash script in the init solves it .
 
You can change the scheduler in run time.

Here are a couple of links to help link 1 and link 2

I have a start up script on the testing machine that changes the scheduler then starts the VM`s
 
Hi everyone,

I know this exists for virtual machines now (including RW distinction, IOPS, and burst) but is there an upcoming solution for Linux Containers (LXC / pct) or a solution available in the command line? Yes, I know the fact that it's a mount point on the FS makes it really challenging....

Thanks!


Tmanok
 
Last edited:
You must log in or register to reply here.
Top Bottom