I have been debugging the application for several weeks now and have not had any luck trapping the leak because it is extermely hard to reproduce.

The reason I kow it is caused by the garbage collector is because I posted an inquiry in this newsgroup long ago about an issue where fbserver.exe was pegging the CPU on my server (Firebird v2). The conclusion was that the garbage collector was responsible for that behavior. When I reverted back to 1.5.3, the server works just fine.


----- Original Message ----
From: Ivan Prenosil <Ivan.Prenosil@...>
To: Firebird-Support <firebird-support@yahoogroups.com>
Sent: Wednesday, January 3, 2007 10:39:50 AM
Subject: Re: [firebird-support] Re: Terminate Long Running Transaction

>>> Is there a way to have the server terminate a long running
>>> transaction?
>>>
>>> I have a multi-threaded, multi-tier application with a FB 1.5.3 DB.
>>> Every once in a while a thread fails / leaks and I would like to find a
>>> way to kill the transaction within the thread. Since my application is
>>> losing it's handle to the thread I am hoping there is a setting on the
>>> server that can help me here.
>>
>> With Classic server you can task kill the corresponding connection,
>> but asides from that no.
>
> What a shame.
>
> Is there any way to specify a setting on the client to disconnect after a certain period of time?
>
> It does not seem logical to me that the garbage collector should wait indefinitely and thus peg the CPU.

1) Fix your applicaion. ("Every once in a while a thread fails / leaks")

If it does not help:

2) Tell us why do you think the problem is caused by garbage collector ?

3) And just to be sure - are you talking about _FB server_ eating CPU,
not client process (e.g. in active wait loop), right ?

Ivan




[Non-text portions of this message have been removed]