Let me throw around a scenario and let's understand how do we handle such scenario before Spring 21.

From Queueable Apex, when we try to make a callout to an external platform, because of a blip in the network, if the callout fails, how do we make sure that the callout it made again and re-enqueued?

We need to get the JobId after the callout is made, then check the status of the Job, if it failed then we need to re-enqueue it manually.

Salesforce team launched something called Finalizers which will make the whole process of re-enqueueing the failed queueable job a breeze.

Let's look at an example!

public class SampleAsyncJob implements Queueable {
    public void execute(QueueableContext ctx) {
        UdpateAWSFinalizer f = new UdpateAWSFinalizer();
        //Business logic
public class UdpateAWSFinalizer implements Finalizer {

    public void execute(FinalizerContext ctx) {
        String reqId = ctx.getRequestId();
        String jobId = Id.valueOf(ctx.getAsyncApexJobId());
        if (ctx.getResult() == ParentJobResult.SUCCESS) {
            System.debug('Parent Queueable (job id: ' + jobId + '): completed successfully!');

        } else { // Queueable failed
            //provide a counter & it should not exceed 5 times
            //re-enqueue the job


This is what we are trying to do in the example.

SampleAsyncJob is a Queueable job which has some business logic associated with it(something like making a callout or etc.)

I created another class that implements Finalizer.

Finalizer is an interface and it overrides execute().

execute() accepts a single parameter of type FinalizerContext

FinalizerContext is a class that provides us access to 4 methods.

  • getRequestId()
  • getAsyncApexJobId()
  • getResult()
  • getException()

In UdpateAWSFinalizer we are trying to check if the status of the parent job is successful or failure.

if (ctx.getResult() == ParentJobResult.SUCCESS) {
} else { // Queueable failed
    //provide a counter

If the status of the parent job is a failure then we have to re-enqueue the business logic again.

There is a limit here as well, we cannot re-enqueue the job beyond 5 times.