Dynamic parameter update is an experimental feature. Service interruption may occur during the update. In most cases, interruption lasts between 5 seconds and 1 minute, depending on the deployment topology and state size.
Traditional parameter updates require restarting a deployment, which causes service interruption, data backtracking delays, and resource consumption spikes. The dynamic parameter update feature sends a REST request to a running deployment, allowing it to reuse the existing JobManager and TaskManagers and apply new parameter values through an in-place restart—or without any restart at all. Combined with resource pre-application and state lazy loading, this reduces service interruption from minutes to seconds.
Supported and non-supported parameters
The following table lists which parameters support dynamic updates and which require a full deployment restart.
| Parameter | Supports dynamic update | Notes |
|---|---|---|
| Parallelism | Yes | Not supported in expert mode (fine-grained resource configuration) |
| Checkpointing Interval | Yes | — |
| Checkpointing Timeout time | Yes | — |
| Min Interval Between Checkpoints | Yes | — |
| All other parameters | No | Modify on the Configuration tab and restart the deployment |
Limitations
-
Dynamic parameter update applies only to running deployments.
-
Only nodes whose parallelism is not explicitly specified can be dynamically rescaled. Nodes with a fixed parallelism are excluded from dynamic scaling because they have scenario-specific constraints—for example, a global operator must have a parallelism of 1, and a Kafka source node must not exceed the number of partitions. To allow dynamic parallelism updates, do not set operator-level parallelism via
DataStream#setParallelismor via the source or sink operator configuration. -
Requires Ververica Runtime (VVR) 8.0.1 or later.
Prerequisites
Before you begin, ensure that you have:
-
A running deployment in Realtime Compute for Apache Flink
-
VVR 8.0.1 or later
Apply a dynamic parameter update
-
Go to the Deployments page.
-
Log on to the Realtime Compute for Apache Flink console.
-
Find the workspace to manage and click Console in the Actions column.
-
In the left-side navigation pane, choose O&M > Deployments. Click the name of the deployment to manage.
-
-
On the Configuration tab, click Edit in the upper-right corner of the Resources or Parameters section.
-
Change the values of the parameters to update, then click Save.
-
In the upper-right corner, click Hot-update.
ImportantThe Hot-update button appears only after you modify at least one of the four supported parameters. If you also changed non-supported parameters, those changes will not take effect until you restart the deployment.
-
In the confirmation dialog, review the information and click OK. After the update starts, an icon indicating dynamic parameter update appears on the deployment.
The following figure compares service interruption time between dynamic parameter update and a traditional restart-based update.
To dynamically scale TaskManagers, change the Parallelism value. For details on calculating the required number of TaskManagers, see Configure job resources.
What's next
-
To update parameters that do not support dynamic updates, modify them on the Configuration tab and restart the deployment. See Configure job deployments and Start a job.
-
To automatically adjust parallelism and resource configurations, use the automatic tuning feature. See Automatic performance tuning.
-
To monitor deployment health and diagnose errors, use intelligent deployment diagnostics. See Perform intelligent job diagnostics.