Vincenzo Gianferrari Pini wrote:
Bernd,
Sorry for the late response, I've been offline for some days.
I'm playing with Postage and 2.3.0rc1: Postage is great! Thanks.
:-) Thank you for trying and giving feedback. There is much room for
improvement and there are probably quite some bugs to be found and removed.
I found a couple of problems:
1) Postage doesn't support SMTP Authentication for "int-ext" users.
Perhaps it is expected, as a "real" config.xml has in any case to be
hacked to run postage, and setting <authRequired> to false is easy. But
insuch case it should be said in the docs.
How can I reproduce this problem? Changing authRequired => true on the
Server does not cause any problems.
2) The generated postage_jvmStatistics.*.csv (and postage_errors.*.csv)
files are empty after a run.
We should at least add a line like "This file is intentionally left
blank" :-)
Both files are always created, even if there is nothing to write to them
while running.
jvmStatistics is only written to when the server and Postage are running
under Java5 since they use latest JVM features.
For more info please see http://wiki.apache.org/james/JamesPostage ,
especially "2. Enable Postage to record memory..."
postage_errors is only written to when errors are recorded which
_cannot_ be related to any of the test mails.
Now a question: the time interval between any two mails generated (based
on count-per-min="n") is constant or is there any probability
distribution function (exponential or other)?
You are referring to configuration elements like these:
<send count-per-min="10" subject="ext2int"
text-size-min="10" text-size-max="1000"
binary-size-min="1" binary-size-max="1000"
/>
The mails are distributed constantly (linear) over time.
This is the most simple behavior and there is plenty of room for
improvement.
And what about the text
and binary sizes?
The sizes of the text and binary MIME parts change randomly for every
new mail. min/max are lower/upper limits for the random function.
Over time, the sizes do not converge against the upper limit and the
limits do not change.
As a bottom line: Currently, the load is constant for each run.
This satisfies the initial use case of having a load spec and try to run
a mail server successfully under that load for some time.
It does not satisfy a stress test use case where you want to see _when_
a server is eventually failing.
Bernd
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
