After problems with connection leak and deadlocks in DBCP we made a decision to replace it with Tomcat JDBC-pool. Of course migration was really simple.
But after deploy it on a production environment I noticed, that load on a server with running two Tomcats increase from 4-4.5 to 5.5. We didn't do anything more, except change of pool. Moreover, performance measured with JMeter decrease by about 5%.
I spent some time to tune pool parameters, but without visible effects. I pasted my current config (from <GlobalNamingResources>
in server.xml
) below:
<Resource name="jdbc/xxxxxx"
auth="Container"
type="javax.sql.DataSource"
factory="org.apache.tomcat.jdbc.pool.DataSourceFactory"
initialSize="10"
maxActive="100"
minIdle="10"
maxIdle="50"
maxWait="10000"
testOnBorrow="true"
testOnReturn="false"
testOnConnect="false"
testWhileIdle="false"
validationQuery="SELECT 1 from dual"
validationInterval="30000"
suspectTimeout="60"
timeBetweenEvictionRunsMillis="30000"
removeAbandonedTimeout="60"
removeAbandoned="true"
logAbandoned="true"
abandonWhenPercentageFull="50"
minEvictableIdleTimeMillis="60000"
jmxEnabled="true"
username="xxxxx"
password="xxxxx"
driverClassName="oracle.jdbc.OracleDriver"
url="jdbc:oracle:oci:xxxxx"/>
FairQueue and PoolSweeperEnabled are true
In Spring applicationContext-jdbc.xml I have only:
<bean id="dataSource" class="org.springframework.jndi.JndiObjectFactoryBean">
<property name="resourceRef">
<value>true</value>
</property>
<property name="jndiName">
<value>java:comp/env/jdbc/PortalDB</value>
</property>
</bean>
What am I doing wrong? I thought, that JDBC_pool should be faster than DBCP out of the box.