Hi guys,
----- Original Message -----
> From time to time I run into a PCP archive with even the latest pmchart where
> scrolling through time seems to make pmchart go into CPU hell. Discussion
> on IRC with nathan helped me track down that it is the Load Avg part of the
> Overview view that is part of the root cause, something to do with discrete
> metrics and hinv.ncpu. If I use the vanilla load average view by itself, it
> works fine.
>
> 09:01 nathans: it'd be good to send a note to the list if you could paul?
> this is libpcp interp.c territory & kenj might have some additional
> insights
>
> I have an archive that exhibits this I can share if it helps, or I can raise
> a bugzilla ticket etc.. ?
>
Should be enough info here already - Ken, I think the root cause here is using
log-once for discrete metrics like hinv.ncpu and then using time windows which
exclude that log-once sample. We end up scanning the window for a value for
the discrete metric IIRC... any way we could make libpcp peek outside the time
window for this case? (not going to be easy, I suspect) Or any alternate way
we could tackle this one? Perhaps put a cap on the scan & returning novalues
sooner than a full time window scan would?
thanks.
--
Nathan
|