task.md

Task Service - Task Documentation

Task Definition

A tasks is described in JSON format and base task (template) definition must be available before a task can be created. Task JSON templates are stored in Mongo database in a collection named taskTemplates. Some task properties are statically predefined in the template, others are populated dynamically by the input request and the output response. Below is an example of task template definition:

{
    "name":"exampleTask",
    "url":"https://jsonplaceholder.typicode.com/todos/1",
    "method":"GET",
    "requestPolicy":"example/example/1.0",
    "responsePolicy":"",
    "finalPolicy":"",
    "cacheNamespace":"login",
    "cacheScope":"user"
}

Tasks are created by using their name attribute. If a task template with the given name does not exist, a task will not be created and an error is returned.

Task Execution

Below is an example of creating a task with the template definition given above:

curl -v -X POST http://localhost:8082/v1/task/exampleTask -d '{"exampleInput":{"test":123}}'

The HTTP request will create a task for asynchronous execution and the JSON object given as input will be used as the body of the task request. The caller will receive immediately the taskID as response, and the result of the asynchronous task execution will be stored in the TSA Cache after the task is completed.

Task Executor Configuration

There are two environment variables that control the level of concurrency of the task executor.

EXECUTOR_WORKERS="5"
EXECUTOR_POLL_INTERVAL="1s"

Poll interval specifies how often the executor will try to fetch a pending task from the queue. After a task is fetched, it is given to one of the workers for execution. With the given example of 1 second (1s), the executor will retrieve 1 task per second at most. If there are multiple instances (pods) of the service, multiply by their number (e.g. 5 pods = 5 tasks per second).

If this is not enough, the poll interval can be decreased, or we can slightly modify the polling function to fetch many tasks at once (and also increase the number of workers).

To learn more about the queue and why we use database as queue see queue

Task Storage

We use MongoDB for tasks storage. There are three Mongo collections with different purpose.

1. taskTemplates

The collection contains predefined task definitions in JSON format. Here are defined what tasks can be created and executed by the service.

2. tasks

The collection contains newly created tasks pending for execution. It acts like a FIFO queue and is used by the task executor to retrieve tasks for workers to execute.

3. tasksHistory

The collection contains successfully completed tasks for results querying, audit, reporting and debugging purposes.