Re: frustrating memory anomaly

On Apr 9, 2005, at 1:55 PM, Sacha Mallais wrote:

> On Apr 9, 2005, at 7:24 am, Anthony Brian Arthur wrote:
>
>> Hi,
>>
>> Configuration:
>>
>> OS - OS X server 10.3.8
>> CPU - dual 1.8
>> RAM - 3 gig
>> WO - 5.2 or greater
>>
>> I have an app that fails in deployment (becoming unavailable in the 
>> middle of a transaction) that does not fail while testing the same 
>> scenario in development.  This happens while downloading a rather 
>> large data set to a file.  The deployed memory settings are -Xms256m 
>> -Xmx1800m.  Now, even though this is a large data set, it is 
>> virtually impossible for it to be greater than 1.8 gig.  This is a 
>> raw row fetch from 1 table, no joins.  Meanwhile the whole database 
>> in an oracle dump is only 2.16 gig.
>>
> Hi Brian,
>
> Does the app crash in the middle of the transaction or does the 
> connection fail?
No, the connection is good.  I am certain of that because several pages 
are traversed before reaching the download.  Also, downloads work 
properly with smaller data sets.  What happens is, shortly after the 
offending download starts, say 15-20 seconds, the browser displays a 
page, all blank except 'Application not available' (or some such) at 
the top.

> If you're running out of memory, I would expect to see a OOM exception 
> in the logs.  Are you seeing that?
No, none of the logs are showing an OOM exception.  That includes the 
Access log, Error log, and the one specific to the app.

> I ask because it sounds like you might be timing out in deployment 
> (because the machine is not local to you, the download takes longer).
>
> Try tweaking the connection timeout param in JavaMon.
This makes sense.  Which field is it -- Lifebeat interval, or some 
custom flag in Additional Args?

Thanks,

-b

>
>
>>   So my question is, how do I know for sure that Monitor is really 
>> claiming the specified java heap settings?  I can see it in the 
>> 'current settings' block on the configuration page, but I am 
>> skeptical for the reason I already outlined.  Any suggestions?
>
> I'm not aware of any way to test the maximum heap size from within 
> Java (or from outside Java, for that matter...).  I personally have 
> used the -Xmx setting (though not with a 1800 MB maximum!  :-) with 
> success.
>
>
> sacha
>
>
> -- 
> Sacha Michel Mallais - 800 lb. gorilla
> Global Village Consulting Inc.:   http://www.global-village.net/
> Bottom posting is your friend:
> http://www.river.com/users/share/etiquette/
>