On Wed, Mar 7, 2012 at 4:20 PM, Steve Holdoway <[email protected]> wrote: > On Wed, 2012-03-07 at 10:12 +1300, Jim Cheetham wrote: >> Swap is overrated. These days we have sufficient RAM to keep the OS
> You're suggesting that it's OK to let the system kill off critical > processes?? That's a bit of a DJB approach imo... I've no idea what DJB says, I stopped listening to him years ago :-) OOM killer has criteria that it uses to identify what process to kill; if it makes a bad decision that's a bug and should be reported/fixed. In any case, I have one VM running for about 5 years now, insufficient RAM for the demands we make of it and insufficient disk for swap, and OOM killer kicks in about once per day on average (judging by my exception reports). Doesn't seem to have hurt it :-) For some other machines we run, the number of processes that are allowed to startup is limited, based on their memory requirements and the physical RAM available. We don't just let the machines spawn randomly based on external requests. These machines don't need swap because we engineer the environment. Sadly engineering is hard, and very few computer people ever do it. > What if that process removes all connectivity with said server? You could argue that an essential business function shouldn't be relying on a single server in the first place, but you do have to choose your Single Point of Failure (errm, Multiple Points of Failures?) to be somewhere. > they're up and running when they're stuffed, so just using inittab is no > good - you end up using ping/pong type programs just to check they > really are up. More overhead, more testing... generally more grief. Newer init's like systemd and upstart handle this thing inside process 1. -jim _______________________________________________ Linux-users mailing list [email protected] http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
