About this guide

This guide covers:

  • Overview of durable and transients job stores
  • Available durable job stores
  • Quartz and Clojure class loaders
  • How to use durable job stores with Quartzite

This work is licensed under a Creative Commons Attribution 3.0 Unported License (including images & stylesheets). The source is available on Github.


During Quartz operation, it has to do some housekeeping: track trigger execution history, trigger and job state, misfires and so on. By default Quartz keeps this information in memory. In some cases it is all you need but has obvious drawbacks:

  • If JVM is stopped, killed or crashes, all state is lost
  • Really large number of jobs and triggers may consume a lot of RAM

To address this issue, Quartz supports durable job stores. They store all the scheduler state, not just information about jobs.

Available Durable Quartz Job Stores

Quartz comes with a JDBC-backed job data store out of the box. There is also a MongoDB-backed one.

For more information, see Quartz documentation on JDBC-backed durable Quartz stores.

Quartz and Clojure Class Loaders

Clojure is a compiled language: code is compiled when it is loaded (usually via clojure.core/require). To dynamically load classes Clojure code compiles to, Clojure uses a special class loader. Since Quartz also uses its own class loader, due to the JVM security model it cannot see job classes Clojure compiler generates.

A solution to this issue is described in the following section.

How to use durable job stores with Quartzite

With a SQL Database

Quartz provides several classes for JDBC-backed job stores. This solution was tested using the JobStoreTX class which should be used in applications that are not running inside an application server like Immutant or JBoss AS.

To allow Quartz to see classes the Clojure compiler generates, we must provide a custom implementation of the ClassLoadHelper class which Quartz uses for discovering classes. A working implementation is below:

// Example package
package oceania.myservice.quartz;

import clojure.lang.DynamicClassLoader;
import org.quartz.spi.ClassLoadHelper;

public class DynamicClassLoadHelper extends DynamicClassLoader implements ClassLoadHelper {
    public void initialize() {}

    public <T> Class<? extends T> loadClass(String name, Class<T> clazz)
        throws ClassNotFoundException {
        return (Class<? extends T>) loadClass(name);

    public ClassLoader getClassLoader() {
        return this;

To configure Quartz to use this class loader, modify your quartz.properties file to include the following lines:


## To use the JobStoreTX with Postgresql include the following:
## (Documentation for other database backends is below)
org.quartz.dataSource.db.URL=<jdbc url string here>
org.quartz.dataSource.db.user=<db user here>
org.quartz.dataSource.db.password=<db password here>


For more configuration properties for the dataSource see Quartz Data Source Config

For jobStore see Quartz Job Store Config

Checkout the jobStore configuration options if you want to utilize a database other than PostgreSQL.

Finally, download the Quartz release from Quartz and navigate to the "docs/dbTables" subdirectory to find the SQL table-creation scripts. Run the script appropriate for your SQL backend and you should be all set!

With MongoDB

A solution to the different class loaders issue is to subclass the job store class and (if it allows this) make it use Clojure's clojure.lang.DynamicClassLoader:

package megacorp.myservice.quartz;

import clojure.lang.DynamicClassLoader;
import com.novemberain.quartz.mongodb.MongoDBJobStore;

public class JobStore extends MongoDBJobStore implements org.quartz.spi.JobStore {
  protected ClassLoader getJobClassLoader() {
    // makes it possible for Quartz to load and instantiate jobs that are defined
    // using defrecord without AOT compilation.
    return new DynamicClassLoader();

and then configure Quartz to use it via the quartz.properties file:

org.quartz.scheduler.instanceName = MyServiceScheduler
org.quartz.threadPool.threadCount = 4

## Use MongoDB-backed store to persistently store information about
# scheduled jobs, triggers and their state.

## Quartz plugins
org.quartz.plugin.triggHistory.class = org.quartz.plugins.history.LoggingTriggerHistoryPlugin
org.quartz.plugin.jobHistory.class = org.quartz.plugins.history.LoggingJobHistoryPlugin

For more information about mixed Clojure/Java projects, see Clojure/Java projects with Leiningen 2+.

Future versions of Quartzite may ship with custom job classes like this out of the box.

Wrapping Up

Quartz provides support for durable stores for its state. Using a durable store is a good idea for availability reasons. Due to specifics of how Clojure compiler and Quartz scheduler work, using durable stores with Quartzite requires a little bit of glue code. Quartzite authors continue looking for a good generic solution.

What version of Quartzite does this guide cover?

This guide covers Quartzite 1.1.x (including beta releases).

comments powered by Disqus