Set default FE queue size config to 1

[eabatalov]

This is much better for slow functions that run for many minutes. With queue size of 2 Server doesn't scale up FEs for such functions. This ends up bottlenecking cluster because allocs wait for many minutes in the queue of size 2 until they can run.

[cursor]

This PR is being reviewed by Cursor Bugbot

Details

Your team is on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle for each member of your team.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

Bug: Test Fails to Validate Function Executor Reuse

The test test_server_uses_the_same_function_executor_if_fe_task_queue_doesnt_overflow no longer validates its intended behavior. Changing _FE_ALLOCATIONS_QUEUE_SIZE to 1 limits it to a single invocation, which prevents it from verifying function executor reuse for multiple tasks when the queue doesn't overflow. This makes the test trivial.

indexify/tests/cli/test_server_fe_scaling.py#L26-L27

indexify/indexify/tests/cli/test_server_fe_scaling.py

Lines 26 to 27 in f942560

	# Server side configuration.
	_FE_ALLOCATIONS_QUEUE_SIZE = 1

indexify/tests/cli/test_server_fe_scaling.py#L68-L91

indexify/indexify/tests/cli/test_server_fe_scaling.py

Lines 68 to 91 in f942560

	graph_name = test_graph_name(self)
	version = str(time.time())
	graph = Graph(
	name=graph_name,
	description="test",
	start_node=TestFunction,
	version=version,
	)
	graph = RemoteGraph.deploy(
	graph=graph, code_dir_path=graph_code_dir_path(__file__)
	)
	invocation_ids: List[str] = []
	for _ in range(_FE_ALLOCATIONS_QUEUE_SIZE):
	invocation_id = graph.call(sleep_secs=0.01)
	invocation_ids.append(invocation_id)
	fe_ids: Set[str] = set()
	for invocation_id in invocation_ids:
	output = wait_function_output(graph, invocation_id, TestFunction.name)
	self.assertEqual(len(output), 1)
	fe_ids.add(output[0])

 

---

[eabatalov]

About the cursor detected bug, yes it's not correct tests anymore but I'd prefer to keep it in case we change the queue setting back.