Hi Marko,
----- Original Message -----
> [...]
> Running logconv.pl on an access log on a busy server without defining the
> start/end time can be a long, CPU consuming task. Thus, we always ask
> metrics for a certain time slice only and calculate the actual metrics since
> the PMDA was started to minimize resource usage and to avoid pmcd(1)
> dropping the PMDA (see http://oss.sgi.com/bugzilla/show_bug.cgi?id=1036). In
> few cases the code could perhaps a bit more Perlish but OTOH TIMTOWTDI.
Looks good to me. Can you take a look over the pcp dev branch code, and
verify everything is in order?
> There is, however, one minor issue currently with the PMDA which I think I
> should mention - it doesn't work :)
Details, details. ;) I'll assume there's nothing we can do about this
and it'll get fixed independently of any PCP use of the tool.
>
> To test this locally one would ideally have 389 DS running but testing
> against a previously captured access log (e.g. from the above RHBZ) should
> help to do basic sanity checking.
>
Test qa/960 exercises the PMDA, optionally using this log now if no other
is available. It all seems to be working nicely for me.
cheers.
--
Nathan
|