Java Persistence API: Controlling the Second-Level-Cache

April 21st, 2014 by

Using the Java Persistence API and a decent persistence provider allows us to configure and fine-tune when and how the second level cache is used in our application.

In the following short examples, we’re going to demonstrate those features written as JUnit test cases and running on a H2 in-memory database.

Persistence Unit Configuration

Persistence Unit Configuration



First of all we need some basic setup to run the following examples .. we need to select a JPA persistence provider and database, create a persistence-unit configuration and an environment to run tests on an in-memory database.

For all readers who have done this a bazillion times before – and I guess they have – please feel free to skip directly to the section “Second Level Cache Modes” or if theory bores you, skip to “Example: Selective Caching“.

Maven Dependencies

I’m using EclipseLink as the persistence provider and a H2 in-memory database for the following examples, so these dependencies are part of my pom.xml.

In addition I’m using JUnit (junit:junit:4.11) and Hamcrest (org.hamcrest:hamcrest-all:1.3) to write the test cases.


Persistence Configuration

The persistence provider is configured in the following persistence.xml put in src/main/resources/META-INF.

Nothing special here excepting the shared-cache-element explained below .. the other configuration is there to specify EclipseLink as designated persistence provider, to use an in-memory H2 database, to create the tables automatically derived from our specified entities and to scan for entities in the current classpath so that we don’t have to write a bunch of class-elements into the persistence descriptor.

<?xml version="1.0" encoding="UTF-8"?>
<persistence xmlns="" xmlns:xsi=""
	xsi:schemaLocation=" persistence_2_1.xsd" version="2.1">
	<persistence-unit name="default" transaction-type="RESOURCE_LOCAL">
			<property name="" value="h2" />
			<property name="javax.persistence.jdbc.driver" value="org.h2.Driver" />
			<property name="javax.persistence.jdbc.url" value="jdbc:h2:mem:testdb;create=true" />
			<property name="javax.persistence.jdbc.user" value="" />
			<property name="javax.persistence.jdbc.password" value="" />
			<property name="javax.persistence.sharedCache.mode" value="ALL"/>
			<property name="eclipselink.ddl-generation" value="create-tables" />
			<property name="eclipselink.logging.level" value="FINE" />

Test Bootstrap

Each test is set up using the following bootstrap code. I’ve ommitted it in the examples below to avoid repetition:

EntityManagerFactory emf;
EntityManager em;
EntityTransaction tx;
public void setup() {
	emf = Persistence.createEntityManagerFactory("default");
	em = emf.createEntityManager();
	tx = em.getTransaction();
public void teardown() {

Second Level Cache Modes

You may control the general cache mode for a persistence unit by adding declaring “shared-cache-element” with one of the following values to your persistence.xml file.

Alternatively you may specify the mode when creating the EntityManagerFactory e.g.:

EntityManagerFactor emf = Persistence.createEntityManagerFactory(
  new Properties().add(
    "javax.persistence.sharedCache.mode", "ENABLE_SELECTIVE")
  • ALL: For this persistence unit, all entity data is stored in the 2nd level cache
  • NONE: No data is cached
  • ENABLE_SELECTIVE: Entities marked with @Cachable are cached (see examples below)
  • DISABLE_SELECTIVE: All entities are cached, excepting entities marked with @Cachable(false) (see examples below)
  • UNSPECIFIED: The caching behaviour is not specified, the persistence provider’s default caching behaviour is used

As I’m using selective caching for the following examples, I’m adding the following declaration to my persistence.xml:


Specify Cache Retrieval and Store Modes

When the second level cache is enabled for a persistence unit, we’re now able fine-tune the caching behaviour by setting retrieval- and store-mode-properties at persistence context level or for each entity-manager operation or query level.

Retrieval Mode

This mode defines how data is read from the cache when using queries or calling the entitymanager’s find method.

We may modify the mode by setting the property javax.persistence.retrieveMode to a value of the enum javax.persistence.CacheRetrieveMode:

  • BYPASS: The cache is bypassed and a call to the database is used to retrieve the data.
  • USE: If the data is available in the cache, it is read from this location, else it is fetched from the database

Example bypassing the cache for a lookup:

EntityManager em=...
Map<String, Object> properties = new HashMap<String, Object>();
properties.put("javax.persistence.cache.retrieveMode", "BYPASS");
Long bookId = 1L;
Book book = em.find(Book.class, bookId, properties);

Store Mode

This mode defines how data is stored in the cache.

We may modify the mode by setting the property javax.persistence.storeMode to a value of the enum javax.persistence.CacheStoreMode:

  • BYPASS: Don’t put anything into the cache
  • REFRESH: Data is put/updated in the cache when read and committed into the database a refresh enforced
  • USE: Data is put/updated in the cache when read and committed into the database

Example: Selective Caching

In this example, we’re going to explore how to achieve selective caching for specific entities using a combination of shared-cache-element = ENABLE_SELECTIVE and the class level annotation @Cachable.

Book Entity

This is our book entity .. nothing special here excepting the enforced caching using java.persistence.Cachable set to true at class level.

package com.hascode.tutorial.jpa_caching.entity;
import javax.persistence.Cacheable;
import javax.persistence.Entity;
import javax.persistence.Id;
public class Book implements Serializable {
	private Long id;
	private String title;
	// constructor, getter, setter ..

Test Case 1: Enforced Caching in Action

We’re persisting two book entities here, force the commit on the entity transaction and get a handle to the cache from the entity-manager-factory afterwards.

As we can see, both entities are present in the cache. Now we’re removing one entity by another from the cache and verify for each that they are not present anymore on the cache.

public void shouldCacheACachableEntity() throws Exception {
	Book book1 = new Book(1L, "Some book");
	Book book2 = new Book(2L, "Another book");
	Cache cache = emf.getCache();
	assertThat(cache.contains(Book.class, 1L), is(true));
	assertThat(cache.contains(Book.class, 2L), is(true));
	Book cachedBook = em.find(Book.class, 1L);
	assertThat(cachedBook, notNullValue());
	cache.evict(Book.class, 1L); // clear one designated book from cache
	assertThat(cache.contains(Book.class, 1L), is(false));
	cache.evict(Book.class); // clear all books from cache
	assertThat(cache.contains(Book.class, 2L), is(false));

Person Entity

Again nothing special here excepting that we’ve set @Cachable to false.

package com.hascode.tutorial.jpa_caching.entity;
public class Person implements Serializable {
	private Long id;
	private String name;
	// constructor, getter, setter ommitted ..

Test Case 2: Enforced Cache Bypass

A similar scenarion as above but this time, the entities are not cached neither after being persisted nor after a search for them.

public void shouldNotCacheAnUncachableEntity() throws Exception {
	Person person1 = new Person(1L, "Lisa");
	Person person2 = new Person(2L, "Tim");
	Cache cache = emf.getCache();
	assertThat(cache.contains(Person.class, 1L), is(false));
	assertThat(cache.contains(Person.class, 2L), is(false));
	Person personFound = em.find(Person.class, 1L);
	assertThat(personFound, notNullValue());

Example: Cache Control on EntityManager Operation

This time we’re going to demonstrate cache control for an entity manager operation.

We’re starting by persisting two book entities, we’re purging the whole cache afterwards and we make sure, that the cache does not contain the first persisted entity.

Afterwards we’re searching for the entity adding a property to the entity manager that forces an update for this entity in the cache.

As we can see, the cache now contains a reference to the entity, so we’re resetting our playing area purging the whole cache again.

We’re running the same search as before but this time we’re feeding the entity manager with a property to bypass the cache.

Finally we’re able to see that the cache has been bypassed as expected and contains no reference to our entity.

public void shouldCacheQuery() throws Exception {
	Book book1 = new Book(3L, "Book 1");
	Book book2 = new Book(4L, "Book 2");
	Cache cache = emf.getCache();
	cache.evictAll(); // clear the whole cache
	assertThat(cache.contains(Book.class, 3L), is(false)); // nothing in the
															// cache
	Map<String, Object> props = new HashMap<String, Object>();
	props.put("javax.persistence.cache.storeMode", "REFRESH");
	em.find(Book.class, 3L, props);
	assertThat(cache.contains(Book.class, 3L), is(true)); // now a cache
															// entry
	cache.evictAll(); // clear cache
	assertThat(cache.contains(Book.class, 3L), is(false));
	props.put("javax.persistence.cache.storeMode", "BYPASS"); // no cache
																// entry
																// written
	em.find(Book.class, 3L, props);
	assertThat(cache.contains(Book.class, 3L), is(false)); // no cache entry
															// as expected

Cleaning the Cache

As we can see in the example above, we have three ways to clear the cache:

  • cache.evict(EntityClass.class, primaryKey): Deletes a particular entity from the cache
  • cache.evict(EntityClass.class): Deletes all instances of a particular class and its subclasses from the cache
  • cache.evictAll: Purges the complete second-level-cache

Tutorial Sources

Please feel free to download the tutorial sources fromĀ my Bitbucket repository, fork it there or clone it using Git:

git clone

Running the Examples

Simply run the test cases using the JUnit Runner or using Maven and the following command:

mvn test

Other JPA Articles

Please feel free to have a look at my other articles about the Java Persistence API:


Article Updates

  • 2015-03-03: Links to my other JPA articles added.

    Tags: , , , , , , ,

    2 Responses to “Java Persistence API: Controlling the Second-Level-Cache”

    1. Binh Thanh Nguyen Says:

      Thanks, nice post

    2. Micha Kops Says:

      Thanks, you’re welcome! :)

    Leave a Reply

    Please note, that no personal information like your IP address is stored and you're not required to enter you real name.

    Comments must be approved before they are published, so please be patient after having posted a comment - it might take a short while.

    Please leave these two fields as-is: