On Thu, Feb 13, 2014 at 4:13 PM, Nathan Scott <nathans@xxxxxxxxxx> wrote:
>
> Wow, that is some neat code - I bet you have no loose papers on your
> desk ;)
Haha. I don't consider myself a neat freak... but now that I think
about it, my desk is pretty neat :)
> - all tested, valgrind used, the class encapsulation looks
> very nice in the end result PMDA, just generally neat.
Great. Was hoping for some feedback on that. I did put in a lot of
work just trying to make the resulting API, and easy to maintain for
implementers (such as Qpid PMDA) without incurring too much
performance overhead. So far, I'm very happy with the result, but
what's neat and readable to me, does not always ring true for others.
> I'm appreciating all the metric help text - it all looks sensible &
> the units and semantics used appear to be spot on. Just good stuff
> all round - again, well done!
Thanks. I can't take too much credit though, the metrics names,
descriptions etc were all exported verbatim (with only very minor
tweaks) from Qpid - ie Qpid's QMFv1 interface exports those metrics,
including descriptions, I just mapped them to PCP types :)
>
> Not sure if you came across dbpmda(1), but this option...
>
> options.add_options()
> ("no-pmda", bool_switch(), "run as a non-PMDA for development");
>
> ... made me wonder. I use either/or in testing (PMDA options and/or
> dbpmda), so having the option is still useful - just mentioning the
> tool in case you'd not come across it yet. Can be run as a regular
> user, so can be simpler (and more targeted) than using pmcd itself.
Excellent. No, I hadn't looked into dbpmda, but it looks quite
useful. The no-pmda option cuts straight to the Qpid code so I can
debug / diagnose the Qpid interface. It looks like dbpmda would make
it easier to debug the PMDA interface on top of that. Will have a
closer look sometime.
>> > Do you want a link
>> > or two on http://oss.sgi.com/projects/pcp/source.html for those trees?
>>
> Done.
Thank you very much! :)
Cheers,
pc.
|