Start a deployment - Realtime Compute for Apache Flink - Alibaba Cloud Documentation Center

After you develop a draft and deploy the draft as a deployment, you must start the deployment on the Deployments page to run the deployment. If you want to resume a deployment that is canceled, you must also start the deployment. After you modify the configuration of a parameter that does not support dynamic update, you must start the deployment to make the modification take effect. This topic describes how to start a deployment.

Prerequisites

A deployment is created. For more information, see Create a deployment.

Limits

You can specify startup options for only streaming deployments.
Only Realtime Compute for Apache Flink that uses Ververica Runtime (VVR) 4.0.11 or later supports state data compatibility check.

Precautions

If a RAM user, RAM role, or another Alibaba Cloud account wants to start a deployment, make sure that the RAM user, RAM role, or the Alibaba Cloud account has the permissions to access the namespace to which the deployment belongs. For more information, see Authorize an account to perform operations in a namespace and Permission management.
If you select Latest State or Specific State when you start a deployment, the system performs a state data compatibility check. If you start a deployment that is incompatible with the state data, the deployment may fail to start or the running result of the deployment may not meet your expectations. Proceed with caution. For more information, see Overview.

Procedure

Go to the Start Job panel.
1. Log on to the console of fully managed Flink by using the member that is assigned the owner role in a namespace.
2. In the top navigation bar, select the namespace from the drop-down list.
3. On the O&M > Deployments page, select STREAM or BATCH from the drop-down list of the deployment type.
Find the desired deployment and click Start in the Actions column.

Optional. In the Start Job panel, configure the startup parameters. This step is required only for streaming deployments.

Initial Mode

If you start a deployment for the first time without the need to reuse states or if a deployment cannot reuse states, select this option. If you select this option, you need to determine whether to enable automatic tuning based on your business requirements.

Strategy

Description

Specify source's start time

If you select Initial Mode, you can select Specify source's start time to specify the start time for reading data.

Only the Simple Log Service and ApsaraMQ for Kafka source table connectors allow you to configure the Specify source's start time parameter in the Start Job panel.

The Specify source's start time parameter that is configured in the Start Job panel takes precedence over the startTime parameter that is configured in the DDL code of a draft.

Note

Not all connectors support the configuration of the startTime parameter. You can view the parameters in the WITH clause of each type of connector to check whether the connector type supports the configuration of the startTime parameter. For more information about the parameters in the WITH clause of a connector type, see the related topic of the specific type of connector, such as Parameters in the WITH clause in Create a Log Service source table.
The startTime parameter takes effect only if a new deployment is started and the startTime parameter is configured. If you start a deployment based on a checkpoint or savepoint, the startTime parameter does not take effect even if this parameter is configured.

Configure Autopilot

If you turn on this switch, you need to configure the Resource Tuning Mode parameter. Valid values:

Autopilot Mode: When the resource usage is low, the system automatically downgrades the resource configuration. When the resource usage reaches a specific threshold, the system automatically upgrades the resource configuration. For more information, see Enable and configure Autopilot.
Scheduled Mode: If you set the Resource Tuning Mode parameter to Scheduled Mode, you must select a scheduled plan from the Existing scheduled plan drop-down list. A plan can contain the mappings between multiple groups of resources and time points. You can configure resources based on the resource usage during each time period. For more information, see Enable and configure scheduled tuning.

Resume Mode

You can select a startup strategy and determine whether to enable automatic tuning based on your business requirements.

Strategy	Description
Latest State	If you want to resume a deployment from the latest state of a savepoint or a checkpoint, select this option. If you select this option to start an SQL deployment, Realtime Compute for Apache Flink detects the changes to the SQL statements, runtime parameter configurations, and engine version of the deployment. If a deployment change is detected, we recommend that you click Click to detect next to State Compatibility to check the compatibility. Then, determine the subsequent actions based on the compatibility check result. For more information about the compatibility check result and suggestions, see Compatibility.
Specific State	If you want to resume a deployment from a savepoint of the current deployment, select this option. For more information about how to create a savepoint, see Status set management.
Other deployment state	If you select this option, you must specify a savepoint of another deployment to resume the current deployment. Only deployments between which the state data is compatible can share savepoints. For more information, see Status set management.
Allow Non-restored State Note This parameter is available only for JAR deployments.	By default, fully managed Flink attempts to match all savepoints with the deployment that is being published. If the status of an operator changes due to the modification of the deployment, the tasks of the deployment may not be restored. If you turn on this switch, fully managed Flink skips the states that do not match the deployment and allows the deployment to start. For more information about Allow Non-Restored State, see Allow None-Restored State.
Configure Autopilot	If you turn on this switch, you need to configure the Resource Tuning Mode parameter. Valid values: Autopilot Mode: When the resource usage is low, the system automatically downgrades the resource configuration. When the resource usage reaches a specific threshold, the system automatically upgrades the resource configuration. For more information, see Enable and configure Autopilot. Scheduled Mode: If you set the Resource Tuning Mode parameter to Scheduled Mode, you must select a scheduled plan from the Existing scheduled plan drop-down list. A plan can contain the mappings between multiple groups of resources and time points. You can configure resources based on the resource usage during each time period. For more information, see Enable and configure scheduled tuning.

In the Start Job panel, click Start.
On the O&M > Deployments page, view the status of the deployment. For more information, see View the status of a deployment.

References

After a deployment is started, you can configure parameters for the deployment. For more information, see Parameters section. You can dynamically update the configurations of specific parameters of a deployment. This reduces the service interruption time caused by the start and cancellation of the deployment. For more information, see Dynamically update the parameter configuration for dynamic scaling.
After a deployment is started, you can use the data lineage feature to track data in the deployment. This way, you can identify issues and evaluate the impact of the issues. For more information, see View data lineage.
You can learn about the enterprise-level state backend storage GeminiStateBackend and the performance comparison between GeminiStateBackend and RocksDBStateBackend. For more information, see GeminiStateBackend.