pcp
[Top] [All Lists]

Re: hotproc rfc

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: hotproc rfc
From: fche@xxxxxxxxxx (Frank Ch. Eigler)
Date: Mon, 12 May 2014 11:18:31 -0400
Cc: Martins Innus <minnus@xxxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <1139662762.4765310.1399862104653.JavaMail.zimbra@xxxxxxxxxx> (Nathan Scott's message of "Sun, 11 May 2014 22:35:04 -0400 (EDT)")
References: <536D28B4.6010504@xxxxxxxxxxx> <1139662762.4765310.1399862104653.JavaMail.zimbra@xxxxxxxxxx>
User-agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/21.4 (gnu/linux)
nathans wrote:

> [...]
> AFAIK, noone else has been looking into it recently.  I have wondered for
> some time whether its functionality could be alternately achieved, using
> cgroups (albeit on Linux only) [...]

I get the impression that cgroups are a hammer for a different &
orthogonal nail of a problem.


> [...] There's also the cgclassify(1) daemon, which could be used to
> provide some of the other pmdahotproc bits, by automatically moving
> processes into a closely-monitored cgroup.  [...]

Sure, if one's willing to let a monitoring tool make observable
administrative changes like that.


For what it's worth, a different way one might this hotproc facility
working would be via pmapi client-side logic, kind of like
derived-metrics.  Let a client periodically poll the metrics needed to
compute the 'hotness' predicates, and update its own indom profiles.
That way, the "hotness" definitions need not be hard-coded &
centralized, nor even require any code/config changes at the server.


- FChE

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