pcp
[Top] [All Lists]

Re: [pcp] hinv.*

To: Michele Baldessari <michele@xxxxxxxxxx>
Subject: Re: [pcp] hinv.*
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed, 29 Oct 2014 21:11:34 -0400 (EDT)
Cc: Mark Goodwin <mgoodwin@xxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <20141028075134.GB6825@xxxxxxxxxxxxxxx>
References: <20141021222355.GC6656@xxxxxxxxxxxxxxx> <573592010.74725901.1413999659304.JavaMail.zimbra@xxxxxxxxxx> <54483934.4060802@xxxxxxxxxx> <20141023072636.GB6158@xxxxxxxxxxxxxxx> <1558759660.1356853.1414474824972.JavaMail.zimbra@xxxxxxxxxx> <20141028075134.GB6825@xxxxxxxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: S6OIVFzvOdJw2NF+z95+K4/jXgA3uA==
Thread-topic: hinv.*
Hi Michele,

----- Original Message -----
> Hi Nathan,
> 
> On Tue, Oct 28, 2014 at 01:40:24AM -0400, Nathan Scott wrote:
> > > The default logging here is "once" for hinv, so that might need tweaking
> > > as
> > > well.
> > 
> > Yeah, hmmm, thats a bit awkward - I don't really want us to have to
> > explicitly list each hinv metric in the logging configs.  Not sure
> > what else we can do though.
> 
> Good point. What would be disadvantage of simply logging hinv.* all the
> time? Too big of an increase in archive files?

Yep, that'd be the downside.

> Or should we aim for a split of the hinv.* hierarchy into "extremely
> unlikely to ever change" and "likely to change" metrics?
> Or we just leave things as is and be it ;)

Ken's pointed out the correct path here - keep the log-once, as-is, and
add in logging on an interval of just those hinv metrics that change.

cheers.

--
Nathan

<Prev in Thread] Current Thread [Next in Thread>