[Data] Fix actor pool scale up logic to consider min_workers

[c21]

Why are these changes needed?

This is to fix the actor pool scale up logic, to consider min_workers (minimal required number of actors).
Previously we don't consider it, so num_total_workers == 0 could happen and lead to ZeroDivisionError: division by zero.

Example stack trace:

  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/torch_iterable_dataset.py", line 18, in __iter__
    yield from it
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/dataset_iterator.py", line 570, in make_generator
    for batch in self.iter_batches(
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/dataset_iterator.py", line 156, in iter_batches
    block_iterator, stats = self._to_block_iterator()
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/dataset_iterator/dataset_iterator_impl.py", line 31, in _to_block_iterator
    block_iterator, stats, executor = ds._plan.execute_to_iterator()
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/plan.py", line 530, in execute_to_iterator
    block_iter = itertools.chain([next(gen)], gen)
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/legacy_compat.py", line 49, in execute_to_legacy_block_iterator
    for bundle in bundle_iter:
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/interfaces.py", line 465, in __next__
    return self.get_next()
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/streaming_executor.py", line 116, in get_next
    raise item
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/streaming_executor.py", line 163, in run
    while self._scheduling_loop_step(self._topology) and not self._shutdown:
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/streaming_executor.py", line 217, in _scheduling_loop_step
    op = select_operator_to_run(
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/streaming_executor_state.py", line 362, in select_operator_to_run
    _try_to_scale_up_cluster(topology, execution_id)
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/streaming_executor_state.py", line 419, in _try_to_scale_up_cluster
    per_task_resource = op.incremental_resource_usage()
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/operators/actor_pool_map_operator.py", line 262, in incremental_resource_usage
    if self._autoscaling_policy.should_scale_up(
  File "/home/ray/anaconda3/lib/python3.9/site-packages/ray/data/_internal/execution/operators/actor_pool_map_operator.py", line 422, in should_scale_up
    and num_running_workers / num_total_workers
ZeroDivisionError: division by zero

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[c21]

Verified this test throws exception without the fix.

[c21]

Merged to master. Checked the failed tests are irrelevant.