StartTask interface

Batch will retry Tasks when a recovery operation is triggered on a Node. Examples of recovery operations include (but are not limited to) when an unhealthy Node is rebooted or a Compute Node disappeared due to host failure. Retries due to recovery operations are independent of and are not counted against the maxTaskRetryCount. Even if the maxTaskRetryCount is 0, an internal retry due to a recovery operation may occur. Because of this, all Tasks should be idempotent. This means Tasks need to tolerate being interrupted and restarted without causing any corruption or duplicate data. The best practice for long running Tasks is to use some form of checkpointing. In some cases the StartTask may be re-run even though the Compute Node was not rebooted. Special care should be taken to avoid StartTasks which create breakaway process or install/launch services from the StartTask working directory, as this will block Batch from being able to re-run the StartTask.

Properties

commandLine

The command line of the StartTask. The command line does not run under a shell, and therefore cannot take advantage of shell features such as environment variable expansion. If you want to take advantage of such features, you should invoke the shell in the command line, for example using "cmd /c MyCommand" in Windows or "/bin/sh -c MyCommand" in Linux. If the command line refers to file paths, it should use a relative path (relative to the Task working directory), or use the Batch provided environment variable (https://docs.microsoft.com/en-us/azure/batch/batch-compute-node-environment-variables).

containerSettings

The settings for the container under which the StartTask runs. When this is specified, all directories recursively below the AZ_BATCH_NODE_ROOT_DIR (the root of Azure Batch directories on the node) are mapped into the container, all Task environment variables are mapped into the container, and the Task command line is executed in the container. Files produced in the container outside of AZ_BATCH_NODE_ROOT_DIR might not be reflected to the host disk, meaning that Batch file APIs will not be able to access those files.

environmentSettings

A list of environment variable settings for the StartTask.

maxTaskRetryCount

The maximum number of times the Task may be retried. The Batch service retries a Task if its exit code is nonzero. Note that this value specifically controls the number of retries. The Batch service will try the Task once, and may then retry up to this limit. For example, if the maximum retry count is 3, Batch tries the Task up to 4 times (one initial try and 3 retries). If the maximum retry count is 0, the Batch service does not retry the Task. If the maximum retry count is -1, the Batch service retries the Task without limit.

resourceFiles

A list of files that the Batch service will download to the Compute Node before running the command line. There is a maximum size for the list of resource files. When the max size is exceeded, the request will fail and the response error code will be RequestEntityTooLarge. If this occurs, the collection of ResourceFiles must be reduced in size. This can be achieved using .zip files, Application Packages, or Docker Containers. Files listed under this element are located in the Task's working directory.

userIdentity

The user identity under which the StartTask runs. If omitted, the Task runs as a non-administrative user unique to the Task.

waitForSuccess

Whether the Batch service should wait for the StartTask to complete successfully (that is, to exit with exit code 0) before scheduling any Tasks on the Compute Node. If true and the StartTask fails on a Node, the Batch service retries the StartTask up to its maximum retry count (maxTaskRetryCount). If the Task has still not completed successfully after all retries, then the Batch service marks the Node unusable, and will not schedule Tasks to it. This condition can be detected via the Compute Node state and failure info details. If false, the Batch service will not wait for the StartTask to complete. In this case, other Tasks can start executing on the Compute Node while the StartTask is still running; and even if the StartTask fails, new Tasks will continue to be scheduled on the Compute Node. The default is true.

Property Details

commandLine

The command line of the StartTask. The command line does not run under a shell, and therefore cannot take advantage of shell features such as environment variable expansion. If you want to take advantage of such features, you should invoke the shell in the command line, for example using "cmd /c MyCommand" in Windows or "/bin/sh -c MyCommand" in Linux. If the command line refers to file paths, it should use a relative path (relative to the Task working directory), or use the Batch provided environment variable (https://docs.microsoft.com/en-us/azure/batch/batch-compute-node-environment-variables).

commandLine: string

Property Value

string

containerSettings

The settings for the container under which the StartTask runs. When this is specified, all directories recursively below the AZ_BATCH_NODE_ROOT_DIR (the root of Azure Batch directories on the node) are mapped into the container, all Task environment variables are mapped into the container, and the Task command line is executed in the container. Files produced in the container outside of AZ_BATCH_NODE_ROOT_DIR might not be reflected to the host disk, meaning that Batch file APIs will not be able to access those files.

containerSettings?: TaskContainerSettings

Property Value

environmentSettings

A list of environment variable settings for the StartTask.

environmentSettings?: EnvironmentSetting[]

Property Value

maxTaskRetryCount

The maximum number of times the Task may be retried. The Batch service retries a Task if its exit code is nonzero. Note that this value specifically controls the number of retries. The Batch service will try the Task once, and may then retry up to this limit. For example, if the maximum retry count is 3, Batch tries the Task up to 4 times (one initial try and 3 retries). If the maximum retry count is 0, the Batch service does not retry the Task. If the maximum retry count is -1, the Batch service retries the Task without limit.

maxTaskRetryCount?: number

Property Value

number

resourceFiles

A list of files that the Batch service will download to the Compute Node before running the command line. There is a maximum size for the list of resource files. When the max size is exceeded, the request will fail and the response error code will be RequestEntityTooLarge. If this occurs, the collection of ResourceFiles must be reduced in size. This can be achieved using .zip files, Application Packages, or Docker Containers. Files listed under this element are located in the Task's working directory.

resourceFiles?: ResourceFile[]

Property Value

userIdentity

The user identity under which the StartTask runs. If omitted, the Task runs as a non-administrative user unique to the Task.

userIdentity?: UserIdentity

Property Value

waitForSuccess

Whether the Batch service should wait for the StartTask to complete successfully (that is, to exit with exit code 0) before scheduling any Tasks on the Compute Node. If true and the StartTask fails on a Node, the Batch service retries the StartTask up to its maximum retry count (maxTaskRetryCount). If the Task has still not completed successfully after all retries, then the Batch service marks the Node unusable, and will not schedule Tasks to it. This condition can be detected via the Compute Node state and failure info details. If false, the Batch service will not wait for the StartTask to complete. In this case, other Tasks can start executing on the Compute Node while the StartTask is still running; and even if the StartTask fails, new Tasks will continue to be scheduled on the Compute Node. The default is true.

waitForSuccess?: boolean

Property Value

boolean