[Nagiosplug-devel] perforrmance data for check_http/check_smtp/check_tcp
Subhendu Ghosh
sghosh at sghosh.org
Fri Aug 8 09:30:09 CEST 2003
Ok - so we stay with gettimeofday and microseconds as integers with
labels.
I would probably add a note to the (developing) guidelines that while we
have numbers in microseconds, the precision is more likely to be less -
probably dependent on OS.
-sg
On Thu, 7 Aug 2003, Karl DeBisschop wrote:
> On Thu, 2003-08-07 at 23:18, Subhendu Ghosh wrote:
> > integers would be nice - is micro seconds a realistic granularity or is
> > one-hundredths better?
>
> us is what gettimeofday returns - its easiest to implement in the
> plugins.
>
> The cuurent plan for perf data has time defaulting to units of seconds
> (s) - otherwise the unit lable is no longer optional. So the both 1.3
> and 1.4 will comply with those (developing) guidelines.
>
> --
> Karl
>
> > either way - integers and minimal backward compatability.
> >
> > -sg
> >
> > On Thu, 7 Aug 2003, Karl DeBisschop wrote:
> >
> > > In the 1.3.0 release, as a straw man for people to comment on, I put in
> > > code for perforance data for a few plugins.
> > >
> > > Where the data was time, I put the value in as seconds like 'time=0.045'
> > >
> > > As we worked out ideas for more complete implementation of perf data in
> > > 1.4.0, in became clear that RRDtool/MRTG type solutions were a notable
> > > part of what people want to do with perf data.
> > >
> > > >From my scan of docs for these tools, there would be less work in the
> > > perf data collecter to report integer us rather than float seconds.
> > >
> > > So I'm proposing to change that in the currently implemented plugins so
> > > the data will be 'time=45123us'
> > >
> > > Does that sound like the right approach to people, or should I preserve
> > > the old format even if it mean a little extara porcessing work.
> > >
> > > --
> > > Karl
> > >
> > >
> > >
>
>
--
More information about the Devel
mailing list