Scheduler Reference Guide
Modern applications often need to run specific tasks periodically.
There are two scheduler extensions in Quarkus.
The quarkus-scheduler extension brings the API and a lightweight in-memory scheduler implementation.
The quarkus-quartz extension implements the API from the quarkus-scheduler extension and contains a scheduler implementation based on the Quartz library.
You will only need quarkus-quartz for more advanced scheduling use cases, such as persistent tasks and clustering.
If you add the quarkus-quartz dependency to your project the lightweight scheduler implementation from the quarkus-scheduler extension is automatically disabled.
|
1. Scheduled Methods
A method annotated with @io.quarkus.scheduler.Scheduled is automatically scheduled for invocation.
A scheduled method must not be abstract or private.
It may be either static or non-static.
A scheduled method can be annotated with interceptor bindings, such as @jakarta.transaction.Transactional and @org.eclipse.microprofile.metrics.annotation.Counted.
If there is a bean class that has no scope and declares at least one non-static method annotated with @Scheduled then @Singleton is used.
|
Furthermore, the annotated method must return void and either declare no parameters or one parameter of type io.quarkus.scheduler.ScheduledExecution.
| The annotation is repeatable so a single method could be scheduled multiple times. |
1.1. Inheritance of metadata
A subclass never inherits the metadata of a @Scheduled method declared on a superclass.
For example, suppose the class org.amce.Foo is extended by the class org.amce.Bar.
If Foo declares a non-static method annotated with @Scheduled then Bar does not inherit the metadata of the scheduled method.
In the following example, the everySecond() method is only invoked upon the instance of Foo.
class Foo {
@Scheduled(every = "1s")
void everySecond() {
// ..do something
}
}
@Singleton
class Bar extends Foo {
}1.2. CDI events
Some CDI events are fired synchronously and asynchronously when specific events occur.
| Type | Event description |
|---|---|
|
An execution of a scheduled job completed successfully. |
|
An execution of a scheduled job completed with an exception. |
|
An execution of a scheduled job was skipped. |
|
The scheduler was paused. |
|
The scheduler was resumed. |
|
A scheduled job was paused. |
|
A scheduled job was resumed. |
1.3. Triggers
A trigger is defined either by the @Scheduled#cron() or by the @Scheduled#every() attribute.
If both are specified, the cron expression takes precedence.
If none is specified, the build fails with an IllegalStateException.
1.3.1. CRON
A CRON trigger is defined by a cron-like expression.
For example "0 15 10 * * ?" fires at 10:15am every day.
@Scheduled(cron = "0 15 10 * * ?")
void fireAt1015AmEveryDay() { }The syntax used in CRON expressions is controlled by the quarkus.scheduler.cron-type property.
The values can be cron4j, quartz, unix and spring.
quartz is used by default.
The cron attribute supports Property Expressions including default values and nested
Property Expressions. (Note that "{property.path}" style expressions are still supported but don’t offer the full functionality of Property Expressions.)
@Scheduled(cron = "${myMethod.cron.expr}")
void myMethod() { }If you wish to disable a specific scheduled method, you can set its cron expression to "off" or "disabled".
myMethod.cron.expr=disabledProperty Expressions allow you to define a default value that is used, if the property is not configured.
0 0 15 ? * MON *@Scheduled(cron = "${myMethod.cron.expr:0 0 15 ? * MON *}")
void myMethod() { }If the property myMethod.cron.expr is undefined or null, the default value (0 0 15 ? * MON *) will be used.
1.3.1.1. Time Zones
The cron expression is evaluated in the context of the default time zone. However, it is also possible to associate the cron expression with a specific time zone.
@Scheduled(cron = "0 15 10 * * ?", timeZone = "Europe/Prague") (1)
void myMethod() { }| 1 | The time zone ID is parsed using java.time.ZoneId#of(String). |
The timeZone attribute supports Property Expressions including default values and nested
Property Expressions.
@Scheduled(cron = "0 15 10 * * ?", timeZone = "${myMethod.timeZone}")
void myMethod() { }1.3.2. Intervals
An interval trigger defines a period between invocations.
The period expression is based on the ISO-8601 duration format PnDTnHnMn.nS and the value of @Scheduled#every() is parsed with java.time.Duration#parse(CharSequence).
However, if an expression starts with a digit and ends with d, P prefix will be added automatically. If the expression only starts with a digit, PT prefix is added automatically.
So for example, 15m can be used instead of PT15M and is parsed as "15 minutes".
@Scheduled(every = "15m")
void every15Mins() { }| A value less than one second may not be supported by the underlying scheduler implementation. In that case a warning message is logged during build and application start. |
The every attribute supports Property Expressions including default values and nested
Property Expressions. (Note that "{property.path}" style expressions are still supported but don’t offer the full functionality of Property Expressions.)
@Scheduled(every = "${myMethod.every.expr}")
void myMethod() { }Intervals can be disabled by setting their value to "off" or "disabled".
So for example a Property Expression with the default value "off" can be used to disable the trigger if its Config Property has not been set.
@Scheduled(every = "${myMethod.every.expr:off}")
void myMethod() { }1.4. Identity
By default, a unique identifier is generated for each scheduled method.
This identifier is used in log messages, during debugging and as a parameter of some io.quarkus.scheduler.Scheduler methods.
Therefore, a possibility to specify an explicit identifier may come in handy.
@Scheduled(identity = "myScheduledMethod")
void myMethod() { }The identity attribute supports Property Expressions including default values and nested
Property Expressions. (Note that "{property.path}" style expressions are still supported but don’t offer the full functionality of Property Expressions.)
@Scheduled(identity = "${myMethod.identity.expr}")
void myMethod() { }1.5. Delayed Start of a Trigger
@Scheduled provides two ways to delay the time a trigger should start firing at.
@Scheduled#delay() and @Scheduled#delayUnit() form the initial delay together.
@Scheduled(every = "2s", delay = 2, delayUnit = TimeUnit.HOUR) (1)
void everyTwoSeconds() { }| 1 | The trigger fires for the first time two hours after the application start. |
| The final value is always rounded to full second. |
@Scheduled#delayed() is a text alternative to the properties above.
The period expression is based on the ISO-8601 duration format PnDTnHnMn.nS and the value is parsed with java.time.Duration#parse(CharSequence).
However, if an expression starts with a digit and ends with d, P prefix will be added automatically. If the expression only starts with a digit, PT prefix is added automatically.
So for example, 15s can be used instead of PT15S and is parsed as "15 seconds".
@Scheduled(every = "2s", delayed = "2h")
void everyTwoSeconds() { }
If @Scheduled#delay() is set to a value greater than zero the value of @Scheduled#delayed() is ignored.
|
The main advantage over @Scheduled#delay() is that the value is configurable.
The delay attribute supports Property Expressions including default values and nested Property Expressions.
(Note that "{property.path}" style expressions are still supported but don’t offer the full functionality of Property Expressions.)
@Scheduled(every = "2s", delayed = "${myMethod.delay.expr}") (1)
void everyTwoSeconds() { }| 1 | The config property myMethod.delay.expr is used to set the delay. |
1.6. Delayed Execution
@Scheduled#executionMaxDelay() can be set to delay each execution of a scheduled method.
The value represents the maximum delay between the activation of the trigger and the execution of the scheduled method.
The actual delay is a randomized number between 0 and the maximum specified delay.
The value is parsed with DurationConverter#parseDuration(String).
It can be a property expression, in which case, the scheduler attempts to use the configured value instead: @Scheduled(executionMaxDelay = "${myJob.maxDelay}").
Additionally, the property expression can specify a default value:
@Scheduled(executionMaxDelay = "${myJob.maxDelay}:500ms}").
@Scheduled(every = "2s", executionMaxDelay = "500ms") (1)
void everyTwoSeconds() { }| 1 | The delay will be a value between 0 and 500 milliseconds. As a result, the period between to everyTwoSeconds() executions will be roughly between one and a half and two and a half seconds. |
1.7. Concurrent Execution
By default, a scheduled method can be executed concurrently.
Nevertheless, it is possible to specify the strategy to handle concurrent executions via @Scheduled#concurrentExecution().
import static io.quarkus.scheduler.Scheduled.ConcurrentExecution.SKIP;
@Scheduled(every = "1s", concurrentExecution = SKIP) (1)
void nonConcurrent() {
// we can be sure that this method is never executed concurrently
}| 1 | Concurrent executions are skipped. |
A CDI event of type io.quarkus.scheduler.SkippedExecution is fired when an execution of a scheduled method is skipped.
|
| Note that only executions within the same application instance are considered. This feature is not intended to work across the cluster. |
1.8. Conditional Execution
You can define the logic to skip any execution of a scheduled method via @Scheduled#skipExecutionIf().
The specified class must implement io.quarkus.scheduler.Scheduled.SkipPredicate and the execution is skipped if the result of the test() method is true.
The class must either represent a CDI bean or declare a public no-args constructor.
In case of CDI, there must be exactly one bean that has the specified class in its set of bean types, otherwise the build fails.
Furthermore, the scope of the bean must be active during execution of the job.
If the scope is @Dependent then the bean instance belongs exclusively to the specific scheduled method and is destroyed when the application is shut down.
class Jobs {
@Scheduled(every = "1s", skipExecutionIf = MyPredicate.class) (1)
void everySecond() {
// do something every second...
}
}
@Singleton (2)
class MyPredicate implements SkipPredicate {
@Inject
MyService service;
boolean test(ScheduledExecution execution) {
return !service.isStarted(); (3)
}
}| 1 | A bean instance of MyPredicate.class is used to evaluate whether an execution should be skipped. There must be exactly one bean that has the specified class in its set of bean types, otherwise the build fails. |
| 2 | The scope of the bean must be active during execution. |
| 3 | Jobs.everySecond() is skipped until MyService.isStarted() returns true. |
Note that this is an equivalent of the following code:
class Jobs {
@Inject
MyService service;
@Scheduled(every = "1s")
void everySecond() {
if (service.isStarted()) {
// do something every second...
}
}
}The main idea is to keep the logic to skip the execution outside the scheduled business methods so that it can be reused and refactored easily.
A CDI event of type io.quarkus.scheduler.SkippedExecution is fired when an execution of a scheduled method is skipped.
|
To skip the scheduled executions while the application is starting up/shutting down, you can make use of the io.quarkus.scheduler.Scheduled.ApplicationNotRunning skip predicate.
|
1.9. Non-blocking Methods
By default, a scheduled method is executed on the main executor for blocking tasks.
As a result, a technology that is designed to run on a Vert.x event loop (such as Hibernate Reactive) cannot be used inside the method body.
For this reason, a scheduled method that returns java.util.concurrent.CompletionStage<Void> or io.smallrye.mutiny.Uni<Void>, or is annotated with @io.smallrye.common.annotation.NonBlocking is executed on the Vert.x event loop instead.
class Jobs {
@Scheduled(every = "1s")
Uni<Void> everySecond() { (1)
// ...do something async
}
}| 1 | The return type Uni<Void> instructs the scheduler to execute the method on the Vert.x event loop. |
1.10. How to use multiple scheduler implementations
In some cases, it might be useful to choose a scheduler implementation used to execute a scheduled method.
However, only one Scheduler implementation is used for all scheduled methods by default.
For example, the quarkus-quartz extension provides an implementation that supports clustering, but it also removes the simple in-memory implementation from the game.
Now, if clustering is enabled then it’s not possible to define a scheduled method that would be executed locally on a single node.
Nevertheless, if you set the quarkus.scheduler.use-composite-scheduler config property to true then a composite Scheduler is used instead.
This means that multiple scheduler implementations are kept running side by side.
Furthermore, it’s possible to choose a specific implementation used to execute a scheduled method using @Scheduled#executeWith().
class Jobs {
@Scheduled(cron = "0 15 10 * * ?") (1)
void fireAt10AmEveryDay() { }
@Scheduled(every = "1s", executeWith = Scheduled.SIMPLE) (2)
void everySecond() { }
}| 1 | If the quarkus-quartz extension is present then this method will be executed with the Quartz-specific scheduler. |
| 2 | If quarkus.scheduler.use-composite-scheduler=true is set then this method will be executed with the simple in-memory implementation provided by the quarkus-scheduler extension. |
2. Scheduler
Quarkus provides a built-in bean of type io.quarkus.scheduler.Scheduler that can be injected and used to pause/resume the scheduler and individual scheduled methods identified by a specific Scheduled#identity().
import io.quarkus.scheduler.Scheduler;
class MyService {
@Inject
Scheduler scheduler;
void ping() {
scheduler.pause(); (1)
scheduler.pause("myIdentity"); (2)
if (scheduler.isRunning()) {
throw new IllegalStateException("This should never happen!");
}
scheduler.resume("myIdentity"); (3)
scheduler.resume(); (4)
scheduler.getScheduledJobs(); (5)
Trigger jobTrigger = scheduler.getScheduledJob("myIdentity"); (6)
if (jobTrigger != null && jobTrigger.isOverdue()){ (7)
// the job is late to the party.
}
}
}| 1 | Pause all triggers. |
| 2 | Pause a specific scheduled method by its identity |
| 3 | Resume a specific scheduled method by its identity |
| 4 | Resume the scheduler. |
| 5 | List all jobs in the scheduler. |
| 6 | Get Trigger metadata for a specific scheduled job by its identity. |
| 7 | You can configure the grace period for isOverdue() with quarkus.scheduler.overdue-grace-period |
A CDI event is fired synchronously and asynchronously when the scheduler or a scheduled job is paused/resumed. The payload is io.quarkus.scheduler.SchedulerPaused, io.quarkus.scheduler.SchedulerResumed, io.quarkus.scheduler.ScheduledJobPaused and io.quarkus.scheduler.ScheduledJobResumed respectively.
|
3. Scheduling Long-Running Tasks
Executing a long-running task might yield a warning message similar to the following:
WARN [io.ver.cor.imp.BlockedThreadChecker] (vertx-blocked-thread-checker) Thread Thread[vert.x-worker-thread-1,5,main] has been blocked for 81879 ms, time limit is 60000 ms: io.vertx.core.VertxException: Thread blockedThis is happening because the default worker thread pool is coming from Vert.x which guards against threads being blocked for far too long.
| The amount of time for which a Vert.x worker thread can be blocked is also configurable. |
Therefore, a proper way to execute long tasks is to offload them from the scheduled method to a custom executor service. Here’s an example of such setup for a long-running task that we do not expect to execute often:
@ApplicationScoped
public class LongRunner implements Runnable {
private ExecutorService executorService;
@PostConstruct
void init() {
executorService = Executors.newThreadPerTaskExecutor(Executors.defaultThreadFactory()); (1)
}
@PreDestroy
void destroy() {
executorService.shutdown(); (2)
}
@Scheduled(cron = "{my.schedule}")
public void update() {
executorService.execute(this); (3)
}
@Override
public void run() { (4)
// perform the actual task here
}
}| 1 | Create a fitting executor. In this case, a new thread is created per scheduled task and stopped once the task finishes. |
| 2 | @PreDestroy callback is used to shut down the executor service. |
| 3 | Scheduled method only delegates the job to the custom executor - this prevent Vert.x thread from being blocked. |
| 4 | The bean implements Runnable, a format we can directly pass to the executor service as a parameter. |
4. Programmatic Scheduling
An injected io.quarkus.scheduler.Scheduler can be also used to schedule a job programmatically.
import io.quarkus.scheduler.Scheduler;
@ApplicationScoped
class MyJobs {
@Inject
Scheduler scheduler;
void addMyJob() { (1)
scheduler.newJob("myJob")
.setCron("0/5 * * * * ?")
.setTask(executionContext -> { (2)
// do something important every 5 seconds
})
.schedule(); (3)
}
void removeMyJob() {
scheduler.unscheduleJob("myJob"); (4)
}
}| 1 | This is a programmatic alternative to a method annotated with @Scheduled(identity = "myJob", cron = "0/5 * * * * ?"). |
| 2 | The business logic is defined in a callback. |
| 3 | The job is scheduled once the JobDefinition#schedule() method is called. |
| 4 | A job that was added programmatically can be also removed. |
By default, the scheduler is not started unless a @Scheduled business method is found. You may need to force the start of the scheduler for "pure" programmatic scheduling via quarkus.scheduler.start-mode=forced.
|
| If the Quartz extension is present and the DB store type is used then it’s not possible to pass a task instance to the job definition and a task class must be used instead. The Quartz API can be also used to schedule a job programmatically. |
In certain cases, a more fine-grained approach might be needed which is why Quarkus also exposes java.util.concurrent.ScheduledExecutorService and java.util.concurrent.ExecutorService that can be injected as CDI beans.
However, these executors are used by other Quarkus extensions and therefore should be approached with caution. Furthermore, users are never allowed to shut these executors down manually.
class JobScheduler {
@Inject
ScheduledExecutorService executor;
void everySecondWithDelay() {
Runnable myRunnable = createMyRunnable();
executor.scheduleAtFixedRate(myRunnable, 3, 1, TimeUnit.SECONDS);
}
}5. Scheduled Methods and Testing
It is often desirable to disable the scheduler when running the tests.
The scheduler can be disabled through the runtime config property quarkus.scheduler.enabled.
If set to false the scheduler is not started even though the application contains scheduled methods.
You can even disable the scheduler for particular Test Profiles.
6. Metrics
Some basic metrics are published out of the box if quarkus.scheduler.metrics.enabled is set to true and a metrics extension is present.
If the Micrometer extension is present, then a @io.micrometer.core.annotation.Timed interceptor binding is added to all @Scheduled methods automatically (unless it’s already present) and a io.micrometer.core.instrument.Timer with name scheduled.methods and a io.micrometer.core.instrument.LongTaskTimer with name scheduled.methods.running are registered. The fully qualified name of the declaring class and the name of a @Scheduled method are used as tags.
If the SmallRye Metrics extension is present, then a @org.eclipse.microprofile.metrics.annotation.Timed interceptor binding is added to all @Scheduled methods automatically (unless it’s already present) and a org.eclipse.microprofile.metrics.Timer is created for each @Scheduled method. The name consists of the fully qualified name of the declaring class and the name of a @Scheduled method. The timer has a tag scheduled=true.
7. OpenTelemetry Tracing
If quarkus.scheduler.tracing.enabled is set to true and the OpenTelemetry extension is present then every job execution, either defined with the @Scheduled annotation or scheduled programmatically, automatically creates a span named after the job’s Identity.
8. Run @Scheduled methods on virtual threads
Methods annotated with @Scheduled can also be annotated with @RunOnVirtualThread.
In this case, the method is invoked on a virtual thread.
The method must return void and your Java runtime must provide support for virtual threads.
Read the virtual thread guide for more details.
9. Configuration Reference
Configuration property fixed at build time - All other configuration properties are overridable at runtime
Configuration property |
Type |
Default |
|---|---|---|
The syntax used in CRON expressions. Environment variable: Show more |
|
|
Scheduled task metrics will be enabled if a metrics extension is present and this value is true. Environment variable: Show more |
boolean |
|
Controls whether tracing is enabled. If set to true and the OpenTelemetry extension is present, tracing will be enabled, creating automatic Spans for each scheduled task. Environment variable: Show more |
boolean |
|
By default, only one Scheduler implementations will be started depending on the value of Environment variable: Show more |
boolean |
|
If schedulers are enabled. Environment variable: Show more |
boolean |
|
Scheduled task will be flagged as overdue if next execution time is exceeded by this period. Environment variable: Show more |
|
|
Scheduler can be started in different modes. By default, the scheduler is not started unless a Environment variable: Show more |
|
|
|
About the Duration format
To write duration values, use the standard You can also use a simplified format, starting with a number:
In other cases, the simplified format is translated to the
|
