Andrus,
Correct I did not have access to the JVM on the production server. Even if I did, I could not predict when it was going to happen. In addition, it *never* occurred on the development server.
The choices were not terribly satisfying: either work through the problem and determine an explicit cause (which could take many hours of work), or do as some middleware-developers suggest and replace your major components with the newest stable versions (i.e. eliminate scores of bugs) and 'get back to us if you still have a problem'. :)
Thanks again,
Joe
On Jun 23, 2010, at 2:06 AM, Andrus Adamchik wrote:
> Glad you figured it out. Tough to diagnose your apps when you don't even have access to the JVM to check the state of the threads...
>
> Andrus
>
> On Jun 22, 2010, at 10:17 PM, Joe Baldwin wrote:
>
>> RE Previous Excessive Load Times (60+ secs) /so-called "Cayenne Connection Pool Errors" /so-called "Memory Management Errors" / etc
>>
>> I wanted to thank everyone for commenting on the array of problems that is summarized above. This is my feedback and what I believe to be the cause of the problem(s).
>>
>> A few months ago, I had noticed a recurring and intermittent problem that showed itself as extremely long load times (sometimes 1-2 mins) even under very light load. After reading through the log files and following suggestions from this forum, there was little change in the problem. Since the webapp was hosted on a production server by commercial company, I had to work closely with the tech support guy who was first convinced that it was a problem in my ISP, then he was convinced it was Cayenne, then he was convinced that it was my code, and so on ...
>>
>> However, this problem never once manifested itself on our development server under any kind of load. So I doubted his analysis.
>>
>> After a long process of testing and elimination, and following increasingly eccentric suggestions by the webhost tech guy, we fired the webhost and moved all of our code to a new webhost who had installed the newest stable versions of Java, Tomcat, and MySQL (something the previous webhost refused to do).
>>
>> As a result, we have not experienced the load-time problem or Connection Pool errors since we changed webhosts. We are using the same code, the same Cayenne configuration and version, and the same connector. It looks like it was either some configuration parameters the webhost tech guy had overlooked or it was the 3 year old version of MySQL that they were running (that I found has documented table locking bugs).
>>
>> Anyway thanks for helping out, this was a difficult problem to crack.
>>
>> Joe
>>
>>
>
This archive was generated by hypermail 2.0.0 : Wed Jun 23 2010 - 18:18:27 UTC