[Bulk] Re: confine user(s) to a core with systemd + gcgroup

[lejeczek]

with new rhel 7 it's not difficult to image common setups 
where... which one is it - multi-user.target or 
network.target is the target and no gui
these naturally most often than not will be 
multicore/multisocket systems
and an admin would like to set a sort of hard ceiling above 
which users running their heavy-numbers-crunching apps could 
not go
with old libcgroup it was fairly straightforward, with 
systemd quite opposite, but I might be wrong, only started 
looking at it all
but if I'm right I would be strange redhat missed/dropped 
all these scenarios where users can be controlled/limited by 
cpusets

any other way to restrict users in/via cgroup cpu? I'm still 
thinking systemd

On 30/07/14 18:13, Tim wrote:
> Just a curiosity question:  If you segregate a user to just one CPU
> core, does that really save a system from being crippled if they max out
> that core, or crash it?  I'd have thought that the rest of the system
> would still be affected to some degree.  At the least, it'd have to be
> completely isolated from the user's core.  And it strikes that me that a
> user using the operating system isn't isolated, they interact with it.
>
>