pcp
[Top] [All Lists]

Re: GSOC Queries : Optimized logging of unchanging performance metrics

To: Martins Innus <minnus@xxxxxxxxxxx>, Vineet Purswani <vineetp@xxxxxxxxxx>
Subject: Re: GSOC Queries : Optimized logging of unchanging performance metrics
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 10 Mar 2016 10:41:46 +1100
Cc: PCP <pcp@xxxxxxxxxxx>, Dave Brolley <brolley@xxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <56E02D96.7070900@xxxxxxxxxxx>
References: <CAJXHLwCXuVHyCzrL6qkzr3tagpBruiBBt19KbkJ=VS6p2wogGg@xxxxxxxxxxxxxx> <56E02D96.7070900@xxxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.1
On 10/03/16 01:05, Martins Innus wrote:
Hi Vineet,

     Thanks for the interest in PCP!
...

G'day Vineet.

Further to Martins email, if you'd really like to understand the problem we're trying to find a better solution for here, ...

1. get a standard PCP installation working, preferably on a Linux system (does not matter what sort of Linux)
2. install the proc PMDA ... something like
        $ cd /var/lib/pcp/pmdas/proc
        $ sudo ./Install
3. then log some of the proc metrics (any or all of the proc.psinfo ones would be fine) for half a dozen 10 second samples ... make sure something in the background is doing the odd fork()
4. now inspect the .meta file that pmlogger created with pmdumplog

All sorts of creative data compression techniques are "in scope" here.

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