[event-hubs] add checkkpointing to migration guide sample (#13016)

sdk/eventhub/event-hubs/migrationguide.md

@@ -94,9 +94,9 @@ This and the need to support improvements to the algorithm used for managing par
 Other noteworthy changes:
 
 - The `send` method on the client that allowed sending single events in each call is removed
- in favor of the `sendBatch` to encourage sending events in batches for better throughput.
+  in favor of the `sendBatch` to encourage sending events in batches for better throughput.
 - The `sendBatch` method on the client has two overloads. One takes an array of events. The other takes an
- object of type `EventDataBatch` that should be created using the `createBatch` method on the client. This object represents the batch and can be safely filled until the maximum size allowed.
+  object of type `EventDataBatch` that should be created using the `createBatch` method on the client. This object represents the batch and can be safely filled until the maximum size allowed.
 
 #### Migrating from `EventHubClient` to `EventHubProducerClient` for sending events
 
@@ -257,7 +257,13 @@ const eph = EventProcessorHost.createFromConnectionString(
 // In V2, you get a single event passed to your callback. If you had asynchronous code running in your callback,
 // it is not awaited before the callback is called for the next event.
 const onMessage = (context, event) => {
-  /** your code here **/
+  /** Your code to process the event here **/
+
+  // Note: EventProcessorHost can invoke the onMessage handler
+  // before the previous onMessage handler invocation has completed.
+  // Special care needs to be taken to ensure that checkpointing happens
+  // on events in order.
+  context.checkpointFromEventData(event);
 };
 
 // This is your error handler for errors occuring when receiving events.
@@ -290,7 +296,16 @@ const subscription = eventHubConsumerClient.subscribe(partitionId, {
   // If your callback is an async function or returns a promise, it will be awaited before the
   // callback is called for the next batch of events.
   processEvents: (events, context) => {
-    /** your code here **/
+    /** Your code to process events here **/
+
+    // The events array could be empty, so only checkpoint if it contained events.
+    if (events.length) {
+      // Save a checkpoint for the last event now that we've processed this batch.
+      // Note: EventHubConsumerClient will wait for `processEvents` to return before
+      // calling it again for the same partition id.
+      // This allows you to safely update your checkpoints while receiving events.
+      await context.updateCheckpoint(events[events.length - 1]);
+    }
   },
 
   // Prior to V5, errors were handled by separate callbacks depending