* [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2007-02-21 01:19]:
> 
> >>   I would really rather not see the setrlimit(2) emulation in the
> >>   resource control framework extended.
> >
> >Why ?
> 
> Setrlimit() suffers from a few restrictions:
> 
>       - it's per process
>       - it can only do a soft and hard limit.
> 
> Per process limits have a drawback in that the actual limit is:
> 
>       - maxprocs * limit
>     
> similarly, I have some issues with "tasks" (same issue as
> processes) and projects (in order to have per-user resource limits
> each user need to be assigned its own project and set of resources)
>
> Looking from what the angle of what I as administrator would want to
> do is:
> 
>       - keep total resource consumption of a zone in check
>         (so the zone.* rctls are good)
> 
>       - defined an upperbound for all users individually
>         (I can't easily do that at the moment, I think)
> 
> as a user, I would be most interested in
> 
>       - preventing huge core dumps
>       - preventing runaways
> 
> for this the task and process limits seem adequate.
> 
> But as an administrator, I really would like to have "user.anything"
> type of rctls.

  The way we were hoping to do this was to have "wildcard" projects, so
  that each user would have a project synthesized automatically upon
  login.  Then all the project controls (and the cascaded task controls)
  apply per-user.

  - Stephen

-- 
Stephen Hahn, PhD  Solaris Kernel Development, Sun Microsystems
[EMAIL PROTECTED]  http://blogs.sun.com/sch/
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code

Reply via email to