Re: phases not completing on poly 2.7.6 on RH8

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Wed Jan 21 2004 - 11:10:48 MST


On Wed, 21 Jan 2004, Daniel Barron wrote:

> I have compiled 2.7.6 on RH8 following the instructions from here:
> http://www.web-polygraph.org/mail-archive/users/200312/0008.html

..

> Anyway, 2.7.6 seemed to compile and install just fine. I have had
> difficulty getting any of the polymix to work but I think the
> problem is identical with those I have got to run at all. Eg
> polymix-2. The phases are not completing see:
>
> 005.01| fyi: phase progress:
> duration: 5.00min goal: 3.00min (166.67% complete)
> xact.count: 20
> fill.size: 53712Byte
> xact.errs.count: 0
> xact.errs.ratio: 0.00
> waiting for WSS to freeze

It is waiting for the working set size (WSS) to freeze, which may take
a long time, depending on your request rate. Please see workload
descriptions on the web site or search for "working set" for details.

> 000.43| i-warm 20 0.40 1618 0.00 0 3
> 000.51| i-warm 20 0.00 -1 -1.00 0 3
> 000.60| i-warm 20 0.00 -1 -1.00 0 3
..
> 005.02| i-warm 20 0.00 -1 -1.00 0 0
> 005.10| i-warm 20 0.00 -1 -1.00 0 0

Also, your request rate is zero after the first 30 seconds of the
test. This most likely indicates a very low peak request rate; warm-up
phase uses 10% of the peak request rate and bursts of requests for
embedded objects may create long periods of inactivity. You may
"never" fill your working set at this rate.

> As you can see the warm phase never completes. Is there something I
> am missing? It's tricky as the docs on the site seem to be for 2.8
> but I can't get that to compile.

You are not using any 2.8 features if you are using standard PolyMix
workloads, so most documents should apply as is or with just cosmetic
changes.

> Same problem in polymix1:

I am not sure what is going there without a complete console log from
both client and server side. If I have to guess, I would say that the
proxy somehow prevents the Polygraph server from communicating phase
information to Polygraph clients (e.g., the proxy may strip some HTTP
headers). This is a wild guess, however. I would need to see both
console logs to make more educated guesses.

> The report generation does not seem to work either but I am assuming
> that is because the first phase never completes so nothing is
> logged.

The first phases of both workloads do not log any stats,
unfortunately. It was a useful feature at the time, but now
Polygraph's reporter can isolate stats based on phases and does not
need that old "feature".

HTH,

Alex.

-- 
Protocol performance, functionality, and reliability testing.
Tools, services, and know-how.
http://www.measurement-factory.com/



This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:27 MST