pcp
[Top] [All Lists]

Re: [pcp] pcp grafana and graphite - How to convert pcp metric values in

To: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Subject: Re: [pcp] pcp grafana and graphite - How to convert pcp metric values into percent
From: Amer Ather <aather@xxxxxxxxxxx>
Date: Wed, 2 Jul 2014 14:51:53 -0700
Cc: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>, Martin Spier <mspier@xxxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=netflix.com; s=google; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; bh=bUfqgJZRiYyABbURClnwDPKlqkC6LYbi6t8W6/TX05Y=; b=hRBLnYAObnczEVtyvdxLdaioBN9rcWDFw5gL6YsmlPuqHguXHcGwkUChFpT3+eYn1V cHE0TZcy1jvSGg+jKMFIDOF30BbDa3I1jIdBDTrub1aiG1lDB3y3y4zXtEqxVhSU1ofz fxqCIrZId+/wbF9QHhmYVioJUr9aEGJ607id0=
In-reply-to: <20140702203904.GA21540@xxxxxxxxxx>
References: <y0mha353ky5.fsf@xxxxxxxx> <y0m8uoh3kf4.fsf@xxxxxxxx> <002d01cf9284$dc8fe570$95afb050$@internode.on.net> <002f01cf9285$f0c63160$d2529420$@internode.on.net> <CAM1aq-HgP5+Tsq_sCLWH6GrLOn8UXQWaEVkVaC1p=TT_kMn6Tg@xxxxxxxxxxxxxx> <20140629220735.GA13993@xxxxxxxxxx> <CAM1aq-Hr2ss+b82p1EMCRTjK_Vsbnde_tKVv53MQfP25p4e-aw@xxxxxxxxxxxxxx> <y0m1tu41lld.fsf@xxxxxxxx> <CAM1aq-EGP8=DGUva+JSk3WMeLRisi0Weg3spN1+SLghgrOKVtQ@xxxxxxxxxxxxxx> <53B46CB4.40409@xxxxxxxxxxxxxxxx> <20140702203904.GA21540@xxxxxxxxxx>
Great information!Â

I somehow misunderstood the scope of derived PCP metrics. It make sense if taking into account PCP design philosophy not to keep monitoring client state information. ÂOur interest in PCP as a on-demand tool was mainly due to this design. We don't want to have another tool on the system collecting system stats and adding overhead. We want to use PCP as a troubleshooting tool where high resolution data collection is triggered on-demand basis to debug the issue.

With PCP graphite integration, we also like to evaluate PCP archives/logging feature as a replacement to sar to log historical data. ÂUsing web graphite client (like grafana), we can build intuitive graphs for data correlation to help us discover anomalies quickly.

Question:
I want to make we are implementing the logic on web client correctly. Let's say PMAPI client is fetching metrics at a second interval. To process it one show do the following:
  • fetch the given metric, say kernel.all.cpu.idle
  • calculate the delta
  • Apply rate conversion. As Nattan pointed out,Âcounters need toÂbe converted into a rate (change-in-value over change-in-time) and Âunits (milli-seconds) converted into a utilization by multiplying by 100Âto produce a percent: "kernel.pct.cpu.idle = 100 * kernel.all.cpu.idle / hinv.ncpu". Metric hinv.ncpu can be cached considering it does n't change.
Thanks for all your help,

Regards,
Amer.




On Wed, Jul 2, 2014 at 1:39 PM, Frank Ch. Eigler <fche@xxxxxxxxxx> wrote:
Hi, Ken -

> [...] Âso the 20th anniversary of "starting" PCP has been and gone.
> [...]

Congrats!

> ... so, over 20 years ago, we made a very early decision to NOT hold
> client state in the collection (server) side of the PMAPI. Â[...] ÂI
> still believe this is semantically the correct thing to do ...
> clients, not pmcd, know what to do with a time series of metric
> observations. [...]

Indeed so. ÂA new issue under contention is whether the client side of
PMAPI - libpcp - could/should have rate-conversion capabilities, so
each client programs wouldn't have to reinvent that wheel.

- FChE



--
Thanks,

Amer Ather
Cloud Performance Engineering
<Prev in Thread] Current Thread [Next in Thread>