tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ben Ricker <>
Subject Re: maxThreads
Date Mon, 18 Aug 2003 18:13:18 GMT
On Mon, 2003-08-18 at 12:50, Geralyn M Hollerman wrote:
> Ben Ricker wrote:
> > Running against a database? Are you pooling the connections? We hit
> > maxthreads when either the DB is messed up (i.e., someone locks a table)
> > or before, when the programmers forgot to run socket_close() on the DB
> > connection, thereby returning it to the pool.
> Why, yes I am going against a database - we could never get Tomcat
> connection pooling working right with it, so we went with just internal
> (Ingres db) connection pooling. I don't know a whole lot about that, as
> that's not the primary app I work on - it's kind of been just "there",
> if you know what I mean, but now that you mention it, I think someone
> might be able to lock a table - I'll have to look into that. Forgive me
> for showing my ignorance, but is socket_close() a Java routine? I don't
> think I've seen that before - we probably call it something different.
> Is it anything like invalidating a session?

socket_close() is a Unix system call...just a bit of jargon on the admin
side. I myself am not a programmer, so I do not know what the exact
method in Java is. 

An easy way to figure out if it is database connections not being closed
is to check the connections from the Tomcat server to the database (if
you are admin) or, if you have the access, check the jdbc (or odbc)
connections used on the database. There should not be more then 10; we
run a DB pool of 10 to handle roughly 100k hits per day. We run an
almost 100% dynamic site and those 10 connections handle the load
admirably. If the number of DB connections grows, we know that either
our load has grown all of a sudden or there is a DB issue. 98% of the
time, it is a DB issue. We have gotten rid of the socket closure issue
as the developers jobs were on the line after they audited the socket
closures that caused a few production outages.


Ben Ricker 

View raw message