Well, if you look at top reporting cpu on a multicore server, some processes do, indeed exceed 100%. I suppose there is an argument to report it either way.
But you answered my question, and I thank you. -AJ Martin Pala <[email protected]> wrote: >Hi, > >the total cpu maximum (summary for all CPU core resources) is 100% in >Monit. > >Single threaded application (such as soffice) which uses 100% of single >CPU will count for 100/number_of_cpus [%] > >For example if you have 8 CPUs/cores: > >- multithreaded application (e.g. MySQL) or application with multiple >child processes (Apache) which which will hog all cpus will show up as >100% cpu usage > >- if single threaded application will hog cpu, it will show as >100/8=12.5% > > >The way how monit counts the cpu usage is similar to how memory and >disk usage is presented by the system: 100% memory usage = summary of >all memory sticks in the system, 100% filesystem usage = summary of >free subdisks capacity (in case of RAID 0/5/6). If we'll go with the >model where every CPU will be 100%, then multi-CPU systems will report >for example 800% CPU usage in above described multithreaded application >case, which will be less natural. > > >Regards, >Martin > > >On Jan 20, 2013, at 5:31 PM, Aaron <[email protected]> wrote: > >> I am having trouble with config for process cpu testing. >> >> I have an occasional runaway soffice process that literally consumes >> 90% cpu until restarted. That would mean it's using one core fully. >> >> In "top" this process doesn't dip under 95%, but my test where >90% >for 4 cycles, restart never fires. >> >> Is monit looking at some other number to base its percentage >calculation? >> >> Thanks for any tips. >> >> -AJ >> -- >> To unsubscribe: >> https://lists.nongnu.org/mailman/listinfo/monit-general > > >-- >To unsubscribe: >https://lists.nongnu.org/mailman/listinfo/monit-general
-- To unsubscribe: https://lists.nongnu.org/mailman/listinfo/monit-general
