Control Execution

Sequential tasks can execute their subtasks in a certain order, defined by steps. By default, sequential tasks execute each subtask in the order they appear. When a subtask fails and enters the FAILED state, the sequential task stops executing and sets the remaining subtask states to SKIPPED. To allow users to model more complex task execution flows, Seele provides two attributes to change the default behavior: when and needs.

when

Accepts a string, with possible values: true or previous.ok. The default value is previous.ok.

When a subtask in a sequential task fails, Seele checks the when value of its successor nodes. By default, previous.ok means that the current node will only continue to run if the predecessor node executes successfully. Setting it to true allows Seele to execute the current node even if the predecessor node fails.

needs

Accepts a string specifying the name of the predecessor task as a subtask in the sequential task.

By default, sequential tasks execute subtasks in the order they are declared. Each subtask's predecessor node is the task immediately preceding it. For example, in the following case, the one, two, and three subtasks will be executed in order, as shown in the diagram below.

steps:
  one: # ...
  two: # ...
  three: # ...

By adding when: one to the three configuration, we change the predecessor node of three from two to one. Now, the execution order of the sequential task changes. It still starts with one, but then concurrently executes two and three. Their relationship is shown in the diagram below.

When used with when, previous.ok points to the state of its predecessor task, i.e., the task specified by when. when can only specify the names of tasks declared before it. For example, in the above case, we cannot use when: three in two.