Hi everybody.
I join this discussion because I'm working with Nikolas on this
project and I'm fighting with this problem as well.
Thank you very much for your answer. However I don't know yet how it
will help us.
Indeed we never use SqlMapClient but only SqlMapClientTemplate and we
don't use transaction either.
I agree with you when you says that we're out of connections but this
is the point. The application behaves as if sessions weren't closed.
We saw on th MySQL server that there was about as many connections as
the maxActive parameter. All the connections were in sleep.
We use commons-dbcp for our datasource (BasicDataSource).
What do you think could bring to this kind of behavior?
Thanks a lot
Stéphane Hanser
Le 11 janv. 08 à 22:39, Clinton Begin a écrit :
It's most likely that somewhere in your code you're not properly
ending a
transaction or closing a session.... Every single SqlMapClient usage
should
be wrapped with something like this (unless you're not using a
transaction
at all):
try {
// start
// work
// commit
} finally {
// end
}
Incidentally, if you want to grab the latest source from the trunk
and build
it (simple to do), I've completely removed all of the pooled
scopes. While
this likely won't solve your problem, the stack trace will likely
point you
to your datasouce and it will appear instead as though you're
running out of
connections (unless you're using our simpledatasource, in which case
the
connections will be forcefully reclaimed and you may never know
about the
problem).
I'd do a thorough check of all SqlMapClient usages and make sure
they're in
try/finally blocks and that there is exactly the same number of
starts as
there are ends... (global text search can usually do this fairly
easily).
Clinton
-----Original Message-----
From: Nikolas Kyriazopoulos Panagiotopoulos [mailto:[EMAIL PROTECTED]
]
Sent: January-11-08 8:41 AM
To: [email protected]
Subject: Re: Ibatis throttle - possible deadlock (ibatis 2.2, 2.3)
[I resend this message because I hadn't confirmed the subscription to
the list before sending. Sorry for all those that might receive it
twice]
Hello,
I would like to bring this: https://issues.apache.org/jira/browse/IBATIS-249
into attention.
It seems that either an old problem is resurging (not really solved),
or that we have some configuration problems, so I would appreciate
your help.
Every hour (approximately), suddenly, without a gradual increase in
thread count etc, a deadlock(?) suddenly rises the charge in our web
application and within seconds, the application is dead.
We tend to see threads like the following through jconsole when this
happens:
Name: TP-Processor431
State: WAITING on [EMAIL PROTECTED]
Total blocked: 4 388 Total waited: 248
Stack trace:
java.lang.Object.wait(Native Method)
java.lang.Object.wait(Object.java:474)
com.ibatis.common.util.Throttle.increment(Throttle.java:70)
com.ibatis.common.util.ThrottledPool.pop(ThrottledPool.java:57)
com
.ibatis
.sqlmap.engine.impl.SqlMapExecutorDelegate.popSession(SqlMapExecut
orDelegate.java:930)
com
.ibatis
.sqlmap.engine.impl.SqlMapSessionImpl.<init>(SqlMapSessionImpl.jav
a:51)
com
.ibatis
.sqlmap.engine.impl.SqlMapClientImpl.getLocalSqlMapSession(SqlMapC
lientImpl.java:258)
The database configuration is like this in our Spring 2.5.1
configuration files, for approximately 1000 connected users:
<bean id="dataSource"
class="org.apache.commons.dbcp.BasicDataSource"
destroy-method="close">
[..]
<property name="maxActive" value="100"/>
<property name="maxIdle" value="30"/>
<property name="maxWait" value="10000"/>
</bean>
