Realtime Compute for Apache Flink:Upgrade the engine version of a deployment

Step 1: Back up a draft and deploy the draft

To ensure the stable running of a deployment, we recommend that you clone the draft of the deployment and upgrade the engine version of the cloned draft.

Step 2: Back up the deployment state

In the left-side navigation pane of the Realtime Compute for Apache Flink console, choose O&M > Deployments. On the Deployments page, click the name of the deployment that you want to manage and view the state set of the deployment on the State tab. For more information, see the "View the state generation overview" section in the Manage a state set topic.

• If your deployment uses stateful computing, you must consider whether the state data can be reused for the deployment after the upgrade.

  Before you upgrade the engine version of your deployment, manually create a savepoint for the deployment. This helps you quickly roll back your deployment if an exception occurs during the upgrade. For more information, see the "Manually create a savepoint" section of the Status set management.

  Important

  • In principle, minor versions of the same major version are compatible with each other. However, the upgrade compatibility between major versions is not guaranteed. If the versions before and after the upgrade are compatible with each other, the new version of the engine can read the savepoints that are generated by the old version of the engine. The old version of the engine may not be able to read the savepoints that are generated by the new version of the engine.

  • Realtime Compute for Apache Flink that uses VVR 6.X or later supports two savepoint formats: native format and standard format. The native format supports faster savepoint generation. The standard format provides better compatibility. Therefore, if you want to upgrade the engine version of a deployment to a later minor version of the same major version, we recommend that you use the native format for savepoint generation. If you want to upgrade the engine version of a deployment to a later major version, we recommend that you use the standard format.

• If your deployment is stateless, proceed to the next step.

(Optional) Step 3: Cancel the deployment

In the left-side navigation pane, choose O&M > Deployments. On the Deployments page, find the deployment you want to manage and click Cancel in the Actions column. For more information, see Cancel a deployment.

If the downstream operators support idempotent write operations or duplicate data is acceptable for your business, you can run the deployment and the backed up deployment at the same time.

Step 4: Start the new deployment

• If the new deployment is stateful, perform the following operations to start the deployment: In the Start Job panel, select Resume Mode, select Specific State, and then select the savepoint that is created for the deployment that you created a backup for.

• If the new deployment is stateless, you can select Initial Mode to start the deployment.

For more information about how to start a deployment, see Start a deployment.

Step 5: Check the online status of the new deployment and drop the deployment for which you created a backup

In most cases, if the deployment starts as expected and the first checkpoint is generated after the upgrade, the upgrade is considered initially successful. We recommend that you observe the correctness of business data to ensure the integrity of the upgrade results.

If the deployment runs as expected and the data correctness is verified, you can manually delete the deployment and the savepoint that is generated before the upgrade. For more information about how to delete a savepoint, see the "Manually delete a specified savepoint" section in Status set management.