Difference between Cancel and Terminate in Workflow

Community Support
1

When workflow is cancelled or terminated, in both the cases, temporalio.exceptions.CancelledError is raised.
What’s the difference between cancel and terminate as in both the cases activity is able to perform a cleanup?

PS: Activity I have implemented are sync python activities, worker is instantiated with activity_executor in order to run non-async activities, and worker graceful timeout is set to a very high duration…

PS: Other post mentions that termination is a hard stop of workflow, but that is not happening in this case
References: Difference between Cancel and Terminate - #2 by maxim

2

Termination is a hard stop of a workflow, not the worker. Please provide more information about what exactly is happening. Workflow history would help.

3

I am sharing activity code

def run(self):
        try:
            for pct in range(0, 100, 1):
                activity.heartbeat(pct)
                # below sleep is similar to grpc/any client/database call
                # PS: time.sleep() is used just to emulate network/blocking calls, we should refrain from using sleep
                time.sleep(0.5)
        except CancelledError as ce:
            LOG.info(f'Received the error, {type(ce)}')
            # reraising the exception
            raise ce

This is how my worker params looks like

Worker(
   client,
   graceful_shutdown_timeout=timedelta(days=30),
   activity_executor=ThreadPoolExecutor(max_workers=10)
)

I did two experiments

  1. Cancel the workflow running this activity.
  2. Terminate the workflow running this activity.

I expected Received the error... log to be printed only in experiment 1.
But even in experiment 2, activity is catching the CancelledError.

4

Terminating workflow (or completing it any other way) cancels all its activities. Cancellation gives the workflow an ability to perform whatever business logic is needed for cleanup.

1 Like