Re: Memory Leak in Polyclients

From: Alex Rousskov <rousskov_at_measurement-factory.com>
Date: Mon, 11 Dec 2006 08:34:25 -0700

On Mon, 2006-12-11 at 00:14 +0530, Jegadeesh wrote:
> Hello all,
>
> We ran the polygraph test on squid with 3 clients . One client is on
> x86 and another two clients are on AMD64 machines. We have configured
> workload files with the following details,
> Object size is 100KB
> request rate is 1200
> time duration 2 hours of each phase.
>
> While we running the test, memory usage is increases gradually and the
> polyclient process stopes after 5 to 6 hours after consuming all the
> available swap+memory.

You may be overloading the bench or the device under test:
http://www.web-polygraph.org/docs/reference/models/traffic.html#Sect:3.3
http://www.web-polygraph.org/mail-archive/users/200303/0006.html
http://www.web-polygraph.org/mail-archive/users/200103/0005.html

.. and many similar references.

> Is there any know memory leak issue there on polygraph code.

Not to my knowledge. There is always a chance for workload-dependent
bugs, but we have been running week-long tests at high request rates
without leaks.

If you are still in doubt, consider generating a test report and posting
workload and console logs so that we can review your results and compare
them with your expectations. You can also try running a best-effort test
to see if you get any "leaks" there (but do not use that for production
testing).

Thank you,

Alex.
Received on Tue Dec 12 2006 - 09:24:57 MST

This archive was generated by hypermail 2.2.0 : Wed Nov 19 2008 - 12:00:10 MST