pcp
[Top] [All Lists]

Re: high frequency metrics collection and trace

To: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: high frequency metrics collection and trace
From: fche@xxxxxxxxxx (Frank Ch. Eigler)
Date: Sat, 15 Dec 2012 21:56:46 -0500
Cc: Jun Wang <junwang123@xxxxxxxxx>, PCP <pcp@xxxxxxxxxxx>
In-reply-to: <1355523024.5399.62.camel@xxxxxxxxxxxxxxxxxxxxxxx> (Ken McDonell's message of "Sat, 15 Dec 2012 09:10:24 +1100")
References: <C1D8F78C-DAEA-4ABB-A12E-D195DEE4BA6B@xxxxxxxxx> <1355523024.5399.62.camel@xxxxxxxxxxxxxxxxxxxxxxx>
User-agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/21.4 (gnu/linux)
Hi, Ken -


kenj wrote:

> [...]
>> Does the PMAPI have to use fetch/pull to collect data with certain
>> fixed frequency? I assume that PMCD cannot push data to the monitoring
>> side via PMAPI. 
>
> There is very little client state maintained at pmcd, so clients are
> free to fetch at whatever frequency they desire and indeed pmlogger and
> pmie collect different groups of metrics at different frequencies.
>
> Pushing from pmcd is nver going to happen (this is not an SNMP trap
> model!).

Might it make sense to add a repeated-pmFetch type of operation to the
wire protocol at some point?  Imagine if the latency/throughput
overhead of requesting the same data metrics over and over again
becomes noticeable?  (From a brief tcpdump experiment, pmval request
traffic is almost as large as pmcd response.)

- FChE

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