On 09/11/2013 10:00 PM, Mark Constable wrote: > But that is the point, every other process, except for the last 1/2 dozen > below, are running at a nice of 0 through to -20 (highest priority). It's > only khugepaged that would be at the same niceness as what I am suggesting > for the courier daemons. "nice -n19" is the lowest priority possible and > "ionice -c3" is for when the block storage device is idle.
Apparently it's been a *long* time since I read the Linux documentation for the scheduler. According to sched-nice-design.txt, a nice +19 process will get 1.5% of the available system resources. So... try it out and see what you get. I doubt it'll do much good. Each SMTP connection should get it's own 1.5% of the system resources, which is probably more than enough given the ratio of CPU speed to most network speeds. The same should be true of courierd and the smtp client module. You could turn down the maximum number of connections and deliveries to limit SMTP even further, but you're placing an upper limit on the service as a whole. If you have a compromised password allowing a spammer to use your service, or if you have a malicious user, they're just going to end up dominating your service and degrading the service available to legitimate users. You're probably a lot better off with pythonfilter's ratelimit, or some derivative, or something similar. ------------------------------------------------------------------------------ How ServiceNow helps IT people transform IT departments: 1. Consolidate legacy IT systems to a single system of record for IT 2. Standardize and globalize service processes across IT 3. Implement zero-touch automation to replace manual, redundant tasks http://pubads.g.doubleclick.net/gampad/clk?id=51271111&iu=/4140/ostg.clktrk _______________________________________________ courier-users mailing list courier-users@lists.sourceforge.net Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-users
