On 02/05/2018 04:02 PM, Mike Small wrote:
At what point does it make sense to go to the cgroup level or even
container level and at what point are traditional Unix abstractions like
process groups and sessions adequate?  If Kent is creating all the
processes himself and they all fall in one process group then kill(2) on
the negative of the process group leader should kill them all.

Even if they are daemons? Also, the parent is (currently) a command line utility that goes away each time it is run.

My use case: I specifically want this command line utility to be minimalist for now, but add features in the future (interrogate the daemons about their state, do something interesting about the ones which report an error condition, maybe shove some into a disabled state) without constantly killing all of the daemons. I want the guts of this software to be able to get ahead of the executive orchestrating it. The result will be a far looser confederacy of processes than we are familiar with in monolithic programs.

-kb
_______________________________________________
Discuss mailing list
[email protected]
http://lists.blu.org/mailman/listinfo/discuss

Reply via email to