The connection timeout issue was a bug, not a feature which could be tuned.
I also read the oracle jdbc docs, and the crux of my point was if you have a
pool of 5 connections all of which die overnight, if you do six database
operations, or retry one six times in the morning does the sixth one work -
meaning the bad connections have been taken out of the pool and new good
ones put in.
I assume you can reproduce this problem in a test environment. What happens
if you restart the database while JRun stays running, does the pool
reconnect or is it hosed? If you cannot get new good connections into the
pool in these cases, then it's clearly a bug in the pooling code. At the
last place I worked we used Informix where under load we'd lose connections,
so this type of retry logic (if connection lost then retry) was a standard
feature built into our application.
Cheers
-----Original Message-----
From: Christopher B. Hamlin [mailto:[EMAIL PROTECTED]]
Sent: Thursday, February 15, 2001 9:57 AM
To: JRun-Talk
Subject: Re: Oracle connection lost every morning.....
Jay Walters wrote:
>
> Where is the not connected to oracle? If your application looks the data
> source in JNDI and asks for a new connection you keep getting not
connected?
> Have you kept trying to connect for a while?
>
> I do seem to remember people complaining about timeouts in the Oracle JDBC
> driver (are you using their latest Type 4 one?) so I am not surprised. I
> would think that the connections in the pool may have timed out, but that
> once you exhaust them (not sure how to do that because the pool needs to
> know they are bad, but perhaps just executing sql on them does that) you
> should get good connections out of the pool. Does this work?
I think the documentation for OracleConnectionCacheImpl says
that it has an event listener hooked to the pooled connection,
and that any fatal errors prompt the cache to remove the
physical connection. I would have thought that the drivers
and the pooled stuff (all the client-side stuff) would talk
together since it's all Oracle (and in same drivers jar, no?).
But I may a bit on the optimistic side (on this).
I don't remember the documentation saying anything at all about
timeouts of any kind. Don't know if that's a good sign, though.
ch
--
Christopher B. Hamlin
American Institute of Physics
Suite 1NO1, 2 Huntington Quadrangle, Melville, NY 11747-4502
Email: [EMAIL PROTECTED] Phone: 516/576-2335 FAX: 516/576-2615
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Structure your ColdFusion code with Fusebox. Get the official book at
http://www.fusionauthority.com/bkinfo.cfm
Archives: http://www.mail-archive.com/[email protected]/
Unsubscribe: http://www.houseoffusion.com/index.cfm?sidebar=lists
