users@glassfish.java.net

domain-restart required - once in 30 minutes? (Re: v2u1: server lag under "load"?)

From: Kristian Rink <kawazu_at_zimmer428.net>
Date: Wed, 7 May 2008 11:03:14 +0200

Am Tue, 6 May 2008 15:19:18 +0200
schrieb Kristian Rink <kawazu_at_zimmer428.net>:

> after goin' productive, I do see quite a strange behaviour in our
> gfv2u1 installation, running a small set of different web
> applications: Usually, the server is incredibly fast, and
> shutdown/startup of the domain also does work reasonably fast.
> However, by now I had to do the fourth domain restart within the last
> hour, and I am not sure whether this is how it is supposed to be.

Back at this, it seems that my migration efforts have come to a
(temporary) grinding halt. To cut things short:

- The domain webapps locked up shortly after I left office yesterday in
the evening and remained unresponsive until I restarted the domain at
around 10:40 pm cest. Things worked fine during the night (when nothing
really happened on the machine).

- In the morning, the "usual behaviour" started again as soon as the
first users logged in to our servers. Same procedure all the time -
things work for a while until all our webapps come to a halt and do not
respond anymore.

- After that (morning work hours) the only way to keep things somehow
working was restarting the domain at least all 30 minutes and constantly
(manually) check whether the webapps still do. In no cases however I do
get any log message explaining me what's possibly wrong.

- Reading through various online ressources, I played around with the
"blocking" option of the http-listener which just made things worse
(slower connections, earlier lockups).

Finally (now), restarting the domain doesn't make my webapps come up
anymore, making the server spill messages regarding "too many open
files" to server.log. This is outrageously strange.

Overally, I am helpless. :/ In my testbed environment, our applications
did pretty well in glassfish, but it seems in load situations (even
though the machine itself doesn't really have load at these times, just
there are some users on the server) things so far go crazy. In order to
regain productivity, I switched back to an emergency tomcat
environment and dumped an issue to the glassfish tracker [1], still
hope to get this resolved somehow. :(

Cheers,
Kristian


[1] https://glassfish.dev.java.net/issues/show_bug.cgi?id=499

-- 
Kristian Rink * http://zimmer428.net * http://flickr.com/photos/z428/
jab: kawazu_at_jabber.ccc.de * icq: 48874445 * fon: ++49 176 2447 2771
"One dreaming alone, it will be only a dream; many dreaming together
is the beginning of a new reality." (Hundertwasser)