pcp
[Top] [All Lists]

Re: [question] PCP UI FrontEnd

To: Aurelien Gonnay <aurelien.gonnay@xxxxxxxxx>
Subject: Re: [question] PCP UI FrontEnd
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Tue, 28 Jul 2015 09:03:01 -0400
Cc: "pcp@xxxxxxxxxxx" <pcp@xxxxxxxxxxx>, TED-DEV-CSP <TED-DEV-CSP@xxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <386270db02804fea9051a255c7db761a@xxxxxxxxxxxxxxxxxxxxxxxxx>
References: <f8fde1169e264460bc2e250d2c9d7df0@xxxxxxxxxxxxxxxxxxxxxxxxx> <y0m1tft2sqt.fsf@xxxxxxxx> <386270db02804fea9051a255c7db761a@xxxxxxxxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.4.2.2i
Hi -

> > [...]  So host names like "foo-bar" will be
> > represented with something like "foo-2E-bar".  It's a necessary evil,
> > considering the need to have a bijective mapping between the two
> > namespaces.

> I understand the need for a bijective encoding, but my eyes are not
> so good at decoding on the fly, and it produces very long names
> (especially with archive names included) making hard to display
> metrics afterwards.

This will be improved shortly with brolley's "open directory as
archive" extensions in libpcp (git://oss.sgi.com/brolley/pcp.git
branch brolley/multi-archive).  With that, code already in pmwebd will
arrange to drop the archive file name parts from the graphite metrics.


> > > [...]
> > > Moreover pcp metric definition are more comprehensive than graphite can 
> > > cope
> > > with, and I'm feeling like we are not making the most out of the collected
> > > metrics.
> > 
> > Sorry, I'm not sure what you mean.  Maybe just that the graphite
> > information is lossy, like no events / strings / metadata being
> > propagated from PCP?  That's true, but somewhat implicit in the use of
> > graphite web interfaces.  (We could do more with graphite "events"
> > though.)

> I meant that indeed, units, discrete / counter / instant semantic,
> metric label are not fully propagated to graphite.  Maybe a dashbord
> builder to generate json for grafana could take care of this and
> display metric units in the legend.

Yes.  Or the server-side-png rendering code (used in the graphite /
graphlot / png-grafana modes) could add such metadata to the legend
imagery.


> [archiving json dashboards] I tried that approach, but found that
> the perf penalty + name mangling make it pretty painful to use pcp.

(Well, we hope to "ease his pain" over time.)


> Besides, most of graphite's server side computation functions do not
> seem supported.

Yes, see <http://oss.sgi.com/bugzilla/show_bug.cgi?id=1094>.


> I thought about using a brokered approach with
> https://github.com/brutasse/graphite-api to implement graphite-api
> and delegate actual data fetching to a either a raw finder based on
> pcpapi or a rest finder calling pmwebapi.  [...]

See also pcp2graphite.


- FChE

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