Hi Michele,
----- Original Message -----
> Hi all,
>
> while for most hinv.* metrics I do not expect any changes over time
> (hinv.pagesize, hinv.ncpu, ...), I was a bit surprised that also
> hinv.cpu.clock seems to not change over time (F21 updated today, here).
> For example:
> $ pmval -h localhost -s 5 hinv.cpu.clock
> metric: hinv.cpu.clock
> host: marquez.int.rhx
> semantics: discrete instantaneous value
> units: count
> samples: 5
> interval: 1.00 sec
>
> cpu0 cpu1 cpu2 cpu3 cpu4
> cpu5 cpu6 cpu7
> 3060. 3060. 1862. 3060. 2926.
> 3060. 1596. 2926.
> 3060. 3060. 1862. 3060. 2926.
> 3060. 1596. 2926.
> 3060. 3060. 1862. 3060. 2926.
> 3060. 1596. 2926.
> 3060. 3060. 1862. 3060. 2926.
> 3060. 1596. 2926.
> 3060. 3060. 1862. 3060. 2926.
> 3060. 1596. 2926.
>
> Since the description implies it should be read from /proc:
> $ pminfo -t -h localhost hinv.cpu.clock
> hinv.cpu.clock [clock rate in Mhz for each CPU as reported by /proc/cpuinfo]
>
> A quick 'egrep "^processor|cpu\ MHz" /proc/cpuinfo' does show that they
> change quite often and all the time.
>
> Is this an issue in the pmda or in the description or have I missed
> something else?
Yeah, I'm finding this a bit odd too - the code over in refresh_proc_cpuinfo
(src/pmdas/linux/proc_cpuinfo.c) certainly *looks* like it should reevaluate
the values on every sample. I can look a bit closer next week if its still
an unsolved mystery then (travelling atm), will see if I can reproduce it.
cheers.
--
Nathan
|