Ryan de Laplante wrote:
>
>>> Short of changing the OS, the workaround and Microsoft patches did
>>> not solve the issue for us.
>>> http://docs.sun.com/app/docs/doc/820-3530/gawlz?a=view
>>
>> Microsoft haven't yet released any patches for that...guess they don't
>> think it's serious enough (LOL) :-)
> I think there are two issues. The one that Sun people have described to
> me requires a reboot. We encounter this every few months. Restarting
> GlassFish doesn't work, the database server can't be accessed anymore
> and restarting it doesn't work either. Only a reboot resolves the problem.
Right, that's the nbpool problem.
>
> The second issue can be resolved by restarting the app server Windows
> service. I wonder if the Microsoft patch solves the first issue.
>
> I also have suspicions that the problem is caused by my own code. You
> might remember that we have a JCA connector, and over 100 Connection
> Pools to communicate with EIS's all over the continent. Originally we
> had the timeout set to 5 minutes, and minimum pool size set to 1. It
> was opening and closing connections a lot, and we had this NP Pool
> problem every couple days. When I changed the timeout to 24 hours, we
> could last a couple weeks before NP Pool problems. I sent you or Sun
> the source code to the JSock library hoping you would be able to confirm
> it is my fault, and what the problem is. The support rep said nobody
> looked at my sources because a different customer provided a full app
> for Sun to reproduce the issue.
Hum....can you send it to me? I will take a look at the Microsoft issue
hasn't been fixed event with the other apps, and I suspect it is juts
because the test case isn't simple enough.
I still am wondering if my own app is
> causing these leaks. It would be good for me and Sun to know this,
> since so few people are reporting the issue. It is possible that all of
> us have our own nio socket code causing these troubles.
I doubt this is caused by NIO...last summer we did an extensive analysis
(with microsoft) and 100% sure NIO has nothing to do, event if your
application leaks file descriptor, the OS IO stack should not get
corrupted like it is right now.
A+
-- Jeanfrancois
>
>
>>>
>>> Use Windows Vista or Windows XP.
>>>
>>
>> Exactly.
> Can you recommend a server OS that you know for sure doesn't have this
> problem? Changing OS's will cost us money and we need to make sure we
> choose one that does not have this problem.
>
>>>> Windows bug. Microsoft has some patches to try and solve it, but
>>>> none of them have worked for us. They also said we are one of only
>>>> two customers who have reported this issue and they are having
>>>> difficulty reproducing it. The other customer recently submitted
>>>> the source to their full app to help Sun and Microsoft reproduce the
>>>> issue.
>>
>> Ha...I didn't know it was your customer :-)
> The other customer I was referring to is a Sun customer. We don't know
> who they are, just that they are experiencing this same issue and have
> been working with Sun and Microsoft longer than us to have the Microsoft
> patch created.
>
>
>>
>> Well, your explanation describe exactly what is the problem :-)
>>
>> A+
>>
>> - Jeanfrancois
>
> Thanks,
> Ryan
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>