[Users] Disk IO priority with multiple VE containers and HDDs.

[antisimus]

Hello,

I have OpenVZ node with multiple disks (12), node has multiple VE
containers (30). Each HDD is shared between 2 - 4 VE.
It usually happens that 1 VE can utilise complete HDD and kill disk
performances for other VE containers located on same HDD.

for example:

HDD-1
VE: 101, 102, 103
HDD-2
VE: 104, 105, 106, 107
...

Now when VE 103 overuses HDD-1 only affected is HDD-1, Having global
IO priority will mean that if HDD-2 containers are idle, HW will think
that overall there is no disk over-usage, because total disk usage of
HDD-1 + HDD-2 is no where near limits, but truth is something else,
VE103 already reached HDD-1 disk IO limit and HW will do nothing to
prevent it from doing that.

IO priority might work on single HDD but with multiple HDDs I doubt it
serves the purpose.

Does anybody know any working solution for this ?

Kind Regards,
VE4FUN