Configure the Task Execution Engine
In Deploy, deployment tasks are executed by dedicated worker instances. A Deploy master generates a deployment plan that contains steps that a Deploy worker's task execution engine will carry out to deploy an application. You can read more about masters and workers here
Tuning the task execution engine
You can tune the Deploy workers' task execution engine with the following settings in deploy-task.yaml:
| Setting | Description | Default |
|---|---|---|
deploy.task.shutdown-timeout | Time to wait for the task engine to shut down. | 1 minute |
deploy.task.max-active-tasks | Maximum number of simultaneous running tasks allowed in the system. If this number is reached, the tasks will appear as QUEUED in the Monitoring section. Each QUEUED task will automatically start after a running task completes. | 100 |
deploy.task.recovery-dir | Name of the directory in XL_DEPLOY_SERVER_HOME where task recovery files are stored. | work |
deploy.task.step.retry-delay | Time to wait before rerunning a step that returned a RETRY exit code. | 5 seconds |
You can configure the thread pool that each worker has available for step execution in deploy-task.yaml:
| Setting | Description | Default |
|---|---|---|
deploy.task.step.execution-threads | Amount of threads in the pool. | 32 |
Threads are shared by all running tasks on a worker; they are not created per deployment.
Task execution example
The following example illustrates how you can adjust the deploy.task.step.execution-threads setting to impact task execution.
This example assumes that no other tasks are active in the system, and uses the out-of-the-box internal worker setup. Note that this is not a production setup. This example is only for illustration purposes.
Assume there is an application that contains six deployables, all of type cmd.Command. Each one is configured with a command to sleep for 15 seconds.
In deploy-task.yaml, set the deploy.step.execution-threads property to 2:
deploy.task.step.execution-threads=2
Restart the Deploy server so the settings take effect.
After the server starts, set up a deployment of the application to an environment. In the Deployment Properties, set the orchestrator to parallel-by-deployed. This ensures that the deployment steps will be executed in parallel. Your deployment will look like:
Click Execute to start the execution. Because the core pool size is 2, only two threads will be created and used for step execution. The Deploy execution engine will start executing two steps and the rest of the steps will be in a queued state:
Because you set the deploy.step.execution-threads property to 2, a maximum of two steps are executed at a time. After the first two steps are executed, the next two steps will be picked for execution until all tasks are complete.