On Wed, 2012-03-21 at 16:24 +0000, Colin Guthrie wrote: > 'Twas brillig, and Robert Fox at 20/03/12 08:13 did gyre and gimble: > > Tried to research it, but it look more like it is for servers than > > desktop. Could someone explain what the value-add is of using this on > > the desktop? > > As used in the context of systemd just a general mechanism for tracking > processes. In the olden days, a service could exec other binaries quite > happily. Killing the primary pid of the service meant that any other > binaries launched were not necessarily killed. This mean that the > service needed to implement it's own accounting and tracking system and > there was no way to properly track to what service a given processes > belongs. > > Now with cgroups, it's possible to fully track all the processes started > by a given service and kill each of them properly when stopping the service. > > In addition it's also possible to identify to which service a given > process belongs. e.g. if something is spinning out of control you can > see it in top and then trace it back to the service itself. > > There are numerous other things that can be done with cgroups such as > imposing various limits on a given service (and any subsequent processes > it spawns). > > > Overall this is an accounting mechanism that is something very much > baked into systemd. It's an integral part of it. It's not something we > could simply "turn off" on desktops vs. servers. > > > However, this isn't really the right place to discuss this topic. > > Col >
Thanks Col - very informative and helpful! Cheers, Robert
