Files
robertvoy-ComfyUI-Distributed/tests/test_image_list_runtime_boundaries.py
T
Robert Wojciechowski 7ed60b93ca feat: dynamically distribute image lists across direct workers (#98)
* feat: add Runpod serverless image collector pools

* fix: prevent worker settings clipping and remove preview copy

* fix: initialize worker settings before mounting the form

* fix: validate complete serverless requests against 10 MB limit

* fix: simplify serverless cards and keep status dots circular

* fix: match serverless endpoint box to master card

* fix: preserve serverless capacity and recovery across job failures

Retain durable reservations until provider termination or explicit recovery, protect unresolved pools and credentials, and add scoped reconciliation. Consolidate lifecycle and secret policies and split prompt/media stages with regression coverage.

* feat: add dynamic per-image list queue with Start and End nodes

* fix(image-list): reject upstream data bypasses around Start

* fix(image-list): make optional End text a connection-only input

* fix(image-list): isolate unavailable workers during preflight

* fix(image-list): bound acknowledgements and configure run deadline

* fix(media-sync): preserve output and temp source roots

* fix(image-list): own job lifecycle and preserve worker preparation

* feat(image-list): let master participate unless orchestrator-only

* feat(image-list)!: replace fixed Start extras with growable value_N slots

Distributed Image List Start no longer has reference_image, text,
strength, seed or per-slot mode dropdowns. It now takes images, mask and
up to 10 generic value_N inputs (IMAGE, MASK, STRING, INT, FLOAT,
BOOLEAN) that grow via native V3 Autogrow, each with a matching output.

- One value is shared by every image; one per image pairs in order; any
  other count fails before dispatch. Only images is split into frames.
- Outputs are declared up front (image, mask, item_index, context,
  value_1..value_10). web/imageListStart.js hides unused value outputs,
  mirrors input types, and follows Autogrow compaction so downstream
  links keep their value.
- Image-list protocol bumped to v2; older workers fail preflight.

Breaking: saved workflows using the old Start inputs must recreate Start.

* fix(serverless): disable Clear saved API key while the worker is enabled

The checkbox read "Clear saved API key (disable pool first)", but the
panel never defines "pool" and saving with it ticked on an enabled worker
was rejected by the backend.

- Disable the checkbox while the worker is enabled, with a tooltip saying
  to turn the worker off first. Also disable it when no key is saved.
- Use "worker" instead of "pool" in user-facing serverless errors, the
  serverless guide and the README entry.

* feat(serverless): replace Concurrent jobs with a single Images per run setting

- Remove the separate Concurrent jobs field; rename Jobs per run to
  Images per run (1-16) with a tooltip explaining that Runpod's max
  workers decides how many GPUs run them.
- Backend uses a fixed per-worker limit of 16 outstanding jobs, so all
  images of a run are submitted at once. Unresolved jobs still count
  against it. The limit is constant, so it never has to change while
  jobs are unresolved.
- Legacy 'concurrency' in saved configs and older browser tabs is
  accepted and ignored/dropped; it no longer triggers the busy guard.
- Revert the enable toggle and show the reason when the backend rejects it.
- Trim a pasted api.runpod.ai/v2/<id> URL to the endpoint ID; plainer
  endpoint-ID error otherwise.
- Confirm before a type change deletes a saved Runpod API key.
- Update docs and tests.
2026-10-13 15:20:05 -07:00

341 lines
17 KiB
Python

"""Exercise production per-item orchestration through controlled provider seams."""
import asyncio
import importlib.util
import json
from pathlib import Path
import sys
import types
import unittest
from unittest.mock import AsyncMock, Mock, patch
import torch
from test_image_list import graph_one_region
ROOT = Path(__file__).resolve().parents[1]
class Response:
def __init__(self, data=None, status=200):
self.data, self.status = data, status
async def __aenter__(self):
return self
async def __aexit__(self, *_args):
return None
async def json(self):
return self.data
def raise_for_status(self):
if self.status >= 400:
raise RuntimeError(f'HTTP {self.status}')
class RuntimeBoundaryTests(unittest.IsolatedAsyncioTestCase):
async def asyncSetUp(self):
name = '_image_list_boundaries'
self.modules = {}
for suffix in ('', '.image_list', '.api', '.api.orchestration', '.utils'):
module = types.ModuleType(name + suffix)
module.__path__ = []
self.modules[name + suffix] = module
self.patch_modules = patch.dict(sys.modules, self.modules)
self.patch_modules.start()
self.addCleanup(self.patch_modules.stop)
spec = importlib.util.spec_from_file_location(name + '.image_list.runtime', ROOT / 'image_list/runtime.py')
self.rt = importlib.util.module_from_spec(spec)
sys.modules[spec.name] = self.rt
spec.loader.exec_module(self.rt)
self.worker = {'id': 'worker', 'host': 'worker', 'type': 'remote', 'port': 8188}
self.config = {'workers': [self.worker], 'settings': {}}
self.prompts, self.uploads, self.deletions, self.cancel_requests = [], [], [], []
self.reject = False
self.interrupted = False
self.completed = True
self.queue = {'queue_running': [], 'queue_pending': []}
self.callback = Mock(side_effect=lambda worker, **_kwargs:
'http://127.0.0.1:8188' if worker['type'] == 'local' else 'http://public-master')
self.sync = AsyncMock(side_effect=self.sync_media)
self.dispatch = AsyncMock(side_effect=self.dispatch_prompt)
self.logs = []
dependencies = {
'.api.orchestration.dispatch': types.SimpleNamespace(
preflight_image_list_workers=AsyncMock(return_value={'worker': True}),
dispatch_worker_prompt=self.dispatch),
'.api.orchestration.prompt_transform': types.SimpleNamespace(
_resolve_direct_workers_for_image_list=lambda config, ids: config['workers']),
'.api.orchestration.media_sync': types.SimpleNamespace(
sync_worker_media=self.sync, fetch_worker_path_separator=AsyncMock(return_value='/'),
convert_paths_for_platform=lambda prompt, separator: prompt),
'.utils.config': types.SimpleNamespace(load_config=lambda: self.config,
get_worker_timeout_seconds=lambda: 60,
get_image_list_job_timeout_seconds=lambda config=None: 0),
'.utils.network': types.SimpleNamespace(build_master_url=lambda **kwargs: 'http://public-master',
build_master_callback_url=self.callback,
build_worker_url=lambda worker, path='': f"http://{worker['id']}{path}",
get_client_session=AsyncMock(return_value=self)),
'.utils.logging': types.SimpleNamespace(log=self.logs.append),
}
for suffix, module in dependencies.items():
sys.modules[name + suffix] = module
self.config_module = dependencies['.utils.config']
# Other legacy test modules replace aiohttp during collection. This
# controlled provider fixture needs only its explicit timeout/error seam.
sys.modules['aiohttp'] = types.SimpleNamespace(
ClientTimeout=lambda **kwargs: kwargs, ClientError=ConnectionError)
comfy = types.ModuleType('comfy')
comfy.model_management = types.SimpleNamespace(processing_interrupted=lambda: self.interrupted)
sys.modules['comfy'] = comfy
sys.modules['comfy.model_management'] = comfy.model_management
async def sync_media(self, worker, prompt):
prompt['7']['inputs']['image'] = 'shared/ref.png'
async def dispatch_prompt(self, worker, prompt, meta, **kwargs):
if self.reject:
raise RuntimeError('prompt rejected')
self.prompts.append((json.loads(json.dumps(prompt)), meta))
inputs = prompt['5']['inputs']
token = inputs['worker_result_token']
item = self.rt.consume_worker_item(token)
self.rt.set_master_result(self.context, inputs['worker_item_index'], token, {'image': item['image']})
return f'prompt-{len(self.prompts)}'
def post(self, url, *, json=None, data=None, **kwargs):
if data is not None:
json = __import__('json').loads(data)
if url.endswith('/distributed/image-list/item'):
self.uploads.append(json['token'])
self.rt.store_worker_item(json['token'], json['item'])
else:
self.cancel_requests.append((url, json))
return Response({})
def delete(self, url, **kwargs):
token = url.rsplit('/', 1)[-1]
self.deletions.append(token)
self.rt.delete_worker_item(token)
return Response({})
def get(self, url, **kwargs):
if url.endswith('/queue'):
return Response(self.queue)
pid = url.rsplit('/', 1)[-1]
return Response({pid: {'status': {'completed': True, 'status_str': 'success'}}}
if self.completed else {})
def create(self, count=2):
prompt = graph_one_region()
prompt['7'] = {'class_type': 'LoadImage', 'inputs': {'image': 'shared/ref.png [output]'}}
prompt['4']['inputs']['reference'] = ['7', 0]
self.metadata = {'nodes': [{'id': 42}], 'extra': {'label': 'generic fixture'}}
self.context = self.rt.create_master_job('3', '5', prompt,
[{'image': torch.zeros((1, 2, 2, 3)), 'index': i} for i in range(count)],
['worker'], 'http://public-master', self.metadata)
return self.context
async def test_each_assignment_reuses_media_paths_with_fresh_tokens_and_metadata(self):
context = self.create()
await self.rt.run_image_list_job(context)
self.assertEqual(self.sync.await_count, 1)
self.assertEqual([p['7']['inputs']['image'] for p, _ in self.prompts],
['shared/ref.png', 'shared/ref.png'])
tokens = [p['5']['inputs']['worker_result_token'] for p, _ in self.prompts]
self.assertEqual(len(set(tokens)), 2)
self.assertEqual([p['5']['inputs']['worker_item_index'] for p, _ in self.prompts], [0, 1])
self.assertEqual([meta for _, meta in self.prompts], [self.metadata, self.metadata])
async def test_annotated_media_roots_survive_windows_prep_and_repeated_items(self):
import tempfile
context = self.create(2)
state = self.rt.get_master_job(context)
state.original_prompt['8'] = {'class_type': 'LoadImage', 'inputs': {'image': 'shared\\ref.png [temp]'}}
state.original_prompt['4']['inputs']['second_reference'] = ['8', 0]
name = '_image_list_boundaries.api.orchestration.media_sync'
spec = importlib.util.spec_from_file_location(name, ROOT / 'api/orchestration/media_sync.py')
assert spec is not None and spec.loader is not None
media = importlib.util.module_from_spec(spec)
sys.modules['_image_list_boundaries.utils.logging'].debug_log = self.logs.append
sys.modules['_image_list_boundaries.utils.trace_logger'] = types.SimpleNamespace(
trace_debug=lambda *args: None, trace_info=lambda *args: None)
spec.loader.exec_module(media)
media.fetch_worker_path_separator = AsyncMock(return_value='\\')
converted = Mock(wraps=media.convert_paths_for_platform)
media.convert_paths_for_platform = converted
uploads = []
async def upload(worker, filename, content, digest, mime, **kwargs):
uploads.append((filename, content))
return True, f'worker-input/{len(uploads)}.png'
media._upload_media_to_worker = AsyncMock(side_effect=upload)
media.sync_worker_media = AsyncMock(wraps=media.sync_worker_media)
with tempfile.TemporaryDirectory(prefix='image-list-media-') as directory:
root = Path(directory)
for kind in ('input', 'output', 'temp'):
path = root / kind / 'shared/ref.png'
path.parent.mkdir(parents=True)
path.write_bytes(kind.encode())
def resolve_annotated(filename):
for kind in ('output', 'temp', 'input'):
suffix = f' [{kind}]'
if filename.endswith(suffix):
return str(root / kind / filename[:-len(suffix)])
return str(root / 'input' / filename)
with patch.dict(sys.modules, {name: media, 'folder_paths': types.SimpleNamespace(
get_annotated_filepath=resolve_annotated)}):
await self.rt.run_image_list_job(context)
self.assertEqual(uploads, [('shared/ref.png [output]', b'output'),
('shared/ref.png [temp]', b'temp')])
media.sync_worker_media.assert_awaited_once()
media.fetch_worker_path_separator.assert_awaited_once()
converted.assert_called_once()
self.assertEqual(converted.call_args.args[1], '\\')
for prompt, _ in self.prompts:
self.assertEqual(prompt['7']['inputs']['image'], 'worker-input/1.png')
self.assertEqual(prompt['8']['inputs']['image'], 'worker-input/2.png')
self.assertEqual(state.original_prompt['7']['inputs']['image'], 'shared/ref.png [output]')
async def test_local_worker_callback_does_not_use_public_master_address(self):
self.worker['type'] = 'local'
await self.rt.run_image_list_job(self.create(1))
self.assertEqual(self.prompts[0][0]['5']['inputs']['master_callback_url'], 'http://127.0.0.1:8188')
self.callback.assert_called()
async def test_remote_worker_uses_topology_callback_helper(self):
await self.rt.run_image_list_job(self.create(1))
self.assertEqual(self.prompts[0][0]['5']['inputs']['master_callback_url'], 'http://public-master')
self.callback.assert_called()
def test_metadata_is_forwarded_into_master_start_prompt(self):
from test_prompt_transform import pt
prompt = graph_one_region()
metadata = {'nodes': [{'id': 1}]}
index = pt.PromptIndex(prompt)
transformed = pt.apply_participant_overrides(
prompt, 'master', ['worker'], {'3': 'run'}, 'http://public-master', False,
index, workflow_meta=metadata)
self.assertEqual(transformed['3']['inputs']['image_list_workflow_meta'], metadata)
async def test_aggregate_input_limit_fails_before_upload(self):
self.rt.MAX_REQUEST_BYTES = 64
with self.assertRaisesRegex(ValueError, 'complete.*request'):
await self.rt.run_image_list_job(self.create(1))
self.assertFalse(self.uploads)
self.assertFalse(self.rt._WORKER_ITEMS)
def test_tensor_and_combined_request_limits_preserve_dtype(self):
item = {'image': torch.zeros((1, 4, 4, 3)), 'index': 0}
encoded = self.rt.encode_item(item)
payload = {'context_id': 'example', 'token': 't' * 48, 'item': encoded}
single = self.rt.encode_request_payload(payload)
self.assertEqual(json.loads(single)['item']['image']['dtype'], 'float32')
self.rt.MAX_REQUEST_BYTES = len(single)
self.assertEqual(self.rt.encode_request_payload(payload), single)
item['values'] = {'value_1': item['image']}
item['mask'] = torch.zeros((1, 4, 4))
payload['item'] = self.rt.encode_item(item)
with self.assertRaisesRegex(ValueError, 'complete.*request'):
self.rt.encode_request_payload(payload)
# Size is checked before allocating/copying a 4K tensor onto the CPU.
with self.assertRaisesRegex(ValueError, 'transport limit'):
self.rt.encode_value(torch.empty((1, 2160, 3840, 3), device='meta'), 'IMAGE')
async def test_rejected_submission_reclaims_uploaded_token_without_prompt_id(self):
self.reject = True
with self.assertRaisesRegex(RuntimeError, 'prompt rejected'):
await self.rt.run_image_list_job(self.create(1))
self.assertEqual(self.deletions, self.uploads)
self.assertFalse(self.rt._WORKER_ITEMS)
self.assertFalse(self.rt._MASTER_JOBS)
self.assertFalse(self.cancel_requests)
async def test_user_interrupt_during_pending_submission_is_bounded(self):
entered = asyncio.Event()
async def never_ack(*args, **kwargs):
entered.set()
await asyncio.Event().wait()
self.dispatch.side_effect = never_ack
self.rt.SUBMISSION_CANCEL_GRACE_SECONDS = 0.02
task = asyncio.create_task(self.rt.run_image_list_job(self.create(2)))
await asyncio.wait_for(entered.wait(), 1)
self.interrupted = True
with self.assertRaises(self.rt.ImageListCancelledError):
await asyncio.wait_for(task, 0.8)
self.assertEqual(len(self.uploads), 1)
self.assertEqual(self.deletions, self.uploads)
self.assertFalse(self.rt._MASTER_JOBS)
self.assertTrue(any('ambiguous' in line for line in self.logs), self.logs)
async def test_owned_running_prompt_survives_cold_load_and_old_hour_cap(self):
self.completed = False
context = self.create(2)
clock = [0.0]
calls = [0]
original_get = self.get
def get(url, **kwargs):
if '/history/' in url:
calls[0] += 1
clock[0] += 3700
self.completed = calls[0] >= 3
if url.endswith('/queue'):
return Response({'queue_running': [[0, 'prompt-1']] if calls[0] == 2 else [],
'queue_pending': [[0, 'prompt-1']] if calls[0] == 1 else []})
return original_get(url, **kwargs)
self.rt.time = types.SimpleNamespace(monotonic=lambda: clock[0])
with patch.object(self, 'get', side_effect=get):
result = await self.rt.run_image_list_job(context)
self.assertEqual(len(result[0]), 2)
self.assertGreaterEqual(clock[0], 7200)
self.assertFalse(self.cancel_requests)
async def test_unrelated_busy_prompt_does_not_refresh_owned_liveness(self):
self.completed = False
clock = [0.0]
original_get = self.get
def get(url, **kwargs):
if '/history/' in url:
clock[0] += 70
if url.endswith('/queue'):
return Response({'queue_running': [[0, 'not-our-prompt']], 'queue_pending': []})
return original_get(url, **kwargs)
self.rt.time = types.SimpleNamespace(monotonic=lambda: clock[0])
with patch.object(self, 'get', side_effect=get):
with self.assertRaisesRegex(TimeoutError, 'liveness'):
await self.rt.run_image_list_job(self.create(1))
self.assertEqual(self.cancel_requests, [
('http://worker/queue', {'delete': ['prompt-1']}),
('http://worker/interrupt', {'prompt_id': 'prompt-1'})])
async def test_failed_queue_cleanup_still_interrupts_and_reclaims_staging(self):
self.completed = False
self.config_module.get_image_list_job_timeout_seconds = lambda config=None: 0.03
original_post = self.post
def post(url, **kwargs):
result = original_post(url, **kwargs)
return Response(status=500) if url.endswith('/queue') else result
with patch.object(self, 'post', side_effect=post):
with self.assertRaisesRegex(TimeoutError, 'total deadline'):
await self.rt.run_image_list_job(self.create(2))
self.assertEqual(self.cancel_requests, [
('http://worker/queue', {'delete': ['prompt-1']}),
('http://worker/interrupt', {'prompt_id': 'prompt-1'})])
self.assertEqual(self.deletions, self.uploads)
self.assertTrue(any('cleanup unconfirmed' in line for line in self.logs), self.logs)
self.assertEqual(self.dispatch.await_count, 1)
async def test_configured_total_deadline_cleans_owned_work(self):
self.completed = False
self.config_module.get_image_list_job_timeout_seconds = lambda config=None: 0.03
self.queue = {'queue_running': [[0, 'prompt-1']], 'queue_pending': []}
with self.assertRaisesRegex(TimeoutError, 'total deadline'):
await asyncio.wait_for(self.rt.run_image_list_job(self.create(2)), 1)
self.assertEqual(len(self.uploads), 1)
self.assertEqual(self.deletions, self.uploads)
self.assertFalse(self.rt._MASTER_JOBS)