users@glassfish.java.net

Re: jstacks of hanged glassfish

From: Oleksiy Stashok <oleksiy.stashok_at_oracle.com>
Date: Wed, 21 Mar 2012 16:37:36 +0100

On 03/21/2012 03:38 PM, forums_at_java.net wrote:
> yes, one of these dumps was caused by overloaded db server. Pool size 20
> should be enough, i just tested Jboss connection pool monitor and max
> watermark after about 1 week of use is 19.
I see.

> but it has also other cause, other then jdbc, because it hanged at
> application not doing jdbc at all. After migration to JBoss i found that
> jboss returns empty pages on conditions where GF hanged. It was
> discovered
> that after placing timeout on AJP module it goes away.
>
> ProxyPass / ajp://127.0.0.1:8009/ ttl=60
>
> my guess is that jboss closed idle connection from Apache, apache
> atempted to
> send new request there and because connection was closed empty page was
> returned.
That's weird, I'd expect that if connection taken from cache failed,
Apache has to create new connection and retry the request again.
But if ttl helped, then I agree, most probably the issue was related to
some idle connection detection logic.

IMO GF 3.1.2 (after applying JK patches) shouldn't have any problems w/
that.

Anyway, thank you for the reports and help.

WBR,
Alexey.