Pooling and XA support in 1.0.0
- Christian Schneider
- Jochen Todea
Changes from 0.9.0
pax-jdbc-pool-* modules prior to version 1.0.0 create a pool and a pool-xa DataSourceFactory service. Unfortunately this confused many users. Another problem was that there was no error in the log when pax-jdbc-config did not find the correct DataSourceFactory. It just did not create a DataSource then.
In pax-jdbc 1.0.0 the pooling modules each expose an OSGi service with interface PooledDataSourceFactory. This service is meant to be used by pax-jdbc-config to wrap the DataSource. So the user always refers to the original DataSourceFactory service in the config and just adds the pool and xa config properties to enable pooling. This allows pax-jdbc-config to report when the pooling support or the TransactionManager is missing. We hope this makes it easier for users to get a fully enabled Datasource working.
There are several pooling modules in pax-jdbc:
| Name | Module | Description |
|---|---|---|
| dbcp2 | pax-jdbc-pool-dbcp | Pooling and JTA support using dbcp2 |
| aries | pax-jdbc-pool-aries | Pooling and JTA support using aries transaction jdbc |
| hikari | pax-jdbc-pool-hikari | Pooling and JTA support using hikari |
How it works
Each pooling module publishes an OSGi service with the interface PooledDataSourceFactory. This service allows to create a pooled and optionally XA enabled DataSource from a normal DataSourceFactory. While it is possible to use the pooling support programmatically the main purpose is to let the config support create pooled DataSources from config.
Keep in mind that the pooled XA DataSource will be published using the javax.sql.DataSource interface. So from the outside it will look like a normal DataSource. Inside it will take care that the connection is added to the TransactionManager as an XA resource.
Additonal config properties to enable pooling and XA
The new pooled DataSourceFactory OSGi service provided by pax-jdbc-pool adds the following properties while keeping the properties of the original DataSourceFactory.
| Key | Value | Default | Description |
|---|---|---|---|
| pool | name of the pooling support to use | null (no pooling) | Decide which pooling support to use. For example use pool=dbcp2 to use the dbcp2 support |
| xa | true / false | false | Set xa=true to auto enlist as an XA resource. You will need a TransactionManager service to make this work |
pax-jdbc-pool-dbcp2 configuration
When calling create you should supply the pooled DataSourceFactory with the same properties as the original one. In addition you can also set pooling properties that start with "pool.". These will be forwarded to the pooling library.
See http://commons.apache.org/proper/commons-dbcp/configuration.html.
| Key | Type | Description |
|---|---|---|
| jdbc.pool.blockWhenExhausted | boolean | Sets whether to block when the borrowObject() method is invoked when the pool is exhausted (the maximum number of "active" objects has been reached) |
| jdbc.pool.lifo | boolean | Sets whether the pool has LIFO (last in, first out) behaviour with respect to idle objects - always returning the most recently used object from the pool, or as a FIFO (first in, first out) queue, where the pool always returns the oldest object in the idle object pool |
| jdbc.pool.maxIdle | int | Returns the cap on the number of "idle" instances in the pool |
| jdbc.pool.maxTotal | int | Sets the cap on the number of objects that can be allocated by the pool (checked out to clients, or idle awaiting checkout) at a given time |
| jdbc.pool.maxWaitMillis | long | Sets the maximum amount of time (in milliseconds) the borrowObject() method should block before throwing an exception when the pool is exhausted and getBlockWhenExhausted() is true |
| jdbc.pool.minEvictableIdleTimeMillis | long | Sets the minimum amount of time an object may sit idle in the pool before it is eligible for eviction by the idle object evictor (if any - see setTimeBetweenEvictionRunsMillis(long)) |
| jdbc.pool.minIdle | int | Sets the target for the minimum number of idle objects to maintain in the pool |
| jdbc.pool.numTestsPerEvictionRun | int | Sets the maximum number of objects to examine during each run (if any) of the idle object evictor thread |
| jdbc.pool.softMinEvictableIdleTimeMillis | long | Sets the minimum amount of time an object may sit idle in the pool before it is eligible for eviction by the idle object evictor (if any - see setTimeBetweenEvictionRunsMillis(long)), with the extra condition that at least minIdle object instances remain in the pool |
| jdbc.pool.testOnBorrow | boolean | Sets whether objects borrowed from the pool will be validated before being returned from the borrowObject() method |
| jdbc.pool.testOnCreate | boolean | Sets whether objects created for the pool will be validated before being returned from the borrowObject() method |
| jdbc.pool.testOnReturn | boolean | Sets whether objects borrowed from the pool will be validated when they are returned to the pool via the returnObject() method |
| jdbc.pool.testWhileIdle | boolean | Returns whether objects sitting idle in the pool will be validated by the idle object evictor (if any - see setTimeBetweenEvictionRunsMillis(long)) |
| jdbc.pool.timeBetweenEvictionRunsMillis | long | Sets the number of milliseconds to sleep between runs of the idle object evictor thread |
The following properties will only work for pax-jdbc-pool >= 0.6.0
| Key | Type | Description | Default |
|---|---|---|---|
| jdbc.factory.cacheState | boolean | If true, the pooled connection will cache the current readOnly and autoCommit settings when first read or written and on all subsequent writes. This removes the need for additional database queries for any further calls to the getter. If the underlying connection is accessed directly and the readOnly and/or autoCommit settings changed the cached values will not reflect the current state. In this case, caching should be disabled by setting this attribute to false | true |
| jdbc.factory.defaultAutoCommit | boolean | The default auto-commit state of connections created by this pool. If not set then the setAutoCommit method will not be called | driver default |
| jdbc.factory.defaultCatalog | String | Sets the default "catalog" setting for borrowed Connections | driver default |
| jdbc.factory.defaultReadOnly | boolean | The default read-only state of connections created by this pool. If not set then the setReadOnly method will not be called. (Some drivers don't support read only mode, ex: Informix) | driver default |
| jdbc.factory.defaultTransactionIsolation | int | The default TransactionIsolation state of connections created by this pool. One of the following: (see javadoc for int values)
| driver default |
| jdbc.factory.maxConnLifetimeMillis | long | The maximum lifetime in milliseconds of a connection. After this time is exceeded the connection will fail the next activation, passivation or validation test. A value of zero or less means the connection has an infinite lifetime. | -1 |
| jdbc.factory.validationQuery | String | The SQL query that will be used to validate connections from this pool before returning them to the caller. If specified, this query MUST be an SQL SELECT statement that returns at least one row. If not specified, connections will be validation by calling the isValid() method. |
Example config for pax-jdbc-pool-dbcp2 together with pax-jdbc-config
Create config in etc/org.ops4j.datasource-test.cfg:
osgi.jdbc.driver.name=H2 pool=dbcp2 xa=true databaseName=test user=sa password= dataSourceName=test2 jdbc.pool.maxTotal=8
Installation in karaf >= 3
Install the features
feature:repo-add mvn:org.ops4j.pax.jdbc/pax-jdbc-features/1.4.4/xml/features feature:install transaction jndi pax-jdbc-h2 pax-jdbc-config pax-jdbc-pool-dbcp2
Check the created DataSource:
karaf@root()> service:list javax.sql.DataSource
If the DataSource is not created the log should tell what is wrong.
Tracking errors
If you see a log message like:
Tracking pooling support with filter (&(objectClass=org.ops4j.pax.jdbc.pool.common.PooledDataSourceFactory)(pool=dbcp2)(xa=true))
This means that pax-jdbc-config found your DataSource config and detected that you require a pooling module dbcp2 and xa support. If the DataSource is not installed then this means that you have either not installed the pax-jdbc-pool-dbcp2 module or the transaction manager bundle.