Files
robertvoy-ComfyUI-Distributed/tests/test_serverless_collector.py
Robert Wojciechowski 6b94242110 feat: add Runpod serverless image Collector pools (#96)
* 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.

* 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 15:19:00 -07:00

203 lines
7.0 KiB
Python

import asyncio
import pytest
import torch
from test_collector_list_inputs import _load_collector_module
class _FakeManager:
def __init__(self, *, pending=False, owned=True):
self.pending = pending
self.owned = owned
self.pending_calls = []
self.owned_calls = []
self.cancel_calls = []
def is_pending(self, collector_job_id, participant_id):
self.pending_calls.append((collector_job_id, participant_id))
return self.pending
def owns(self, collector_job_id, participant_id):
self.owned_calls.append((collector_job_id, participant_id))
return self.owned
async def cancel(self, worker_id=None, execution_id=None, collector_job_id=None):
self.cancel_calls.append({
"worker_id": worker_id,
"execution_id": execution_id,
"collector_job_id": collector_job_id,
})
return 1
def _set_up_collector(module, job_id, queue, manager=None):
module.prompt_server.distributed_jobs_lock = asyncio.Lock()
module.prompt_server.distributed_pending_jobs = {job_id: queue}
if manager is not None:
module.prompt_server.distributed_serverless = manager
return module.DistributedCollectorNode()
def test_serverless_pending_waits_past_normal_worker_timeout():
module = _load_collector_module()
job_id = "collector-pending"
participant = "serverless:pool:1"
queue = asyncio.Queue()
manager = _FakeManager(pending=True)
collector = _set_up_collector(module, job_id, queue, manager)
module.get_worker_timeout_seconds = lambda: 0.01
module.HEARTBEAT_INTERVAL = 0.1
probe_calls = []
module.probe_worker = lambda *args, **kwargs: probe_calls.append(args)
async def deliver_after_provider_work():
await asyncio.sleep(0.14)
queue.put_nowait({
"worker_id": participant,
"image_index": 0,
"tensor": torch.ones(1, 2, 2, 2),
"is_last": True,
})
async def run_collection():
producer = asyncio.create_task(deliver_after_provider_work())
result = await collector.execute(
images=None,
audio=None,
multi_job_id=job_id,
enabled_worker_ids=f'["{participant}"]',
delegate_only=True,
)
await producer
return result
images, _audio = asyncio.run(run_collection())
assert tuple(images.shape) == (1, 2, 2, 2)
assert manager.pending_calls
assert probe_calls == []
assert job_id not in module.prompt_server.distributed_pending_jobs
def test_serverless_queue_error_raises_and_cancels_only_collector_binding():
module = _load_collector_module()
job_id = "collector-error"
queue = asyncio.Queue()
queue.put_nowait({
"worker_id": "serverless:pool:1",
"serverless_error": "Provider execution failed safely",
"is_last": True,
})
manager = _FakeManager()
collector = _set_up_collector(module, job_id, queue, manager)
with pytest.raises(RuntimeError, match="serverless.*Provider execution failed safely"):
asyncio.run(collector.execute(
images=None,
audio=None,
multi_job_id=job_id,
enabled_worker_ids='["serverless:pool:1"]',
delegate_only=True,
))
assert manager.cancel_calls == [{
"worker_id": None,
"execution_id": None,
"collector_job_id": job_id,
}]
assert job_id not in module.prompt_server.distributed_pending_jobs
def test_owned_serverless_slot_without_pending_result_fails_instead_of_partial_success():
module = _load_collector_module()
job_id = "collector-stopped"
queue = asyncio.Queue()
manager = _FakeManager(pending=False, owned=True)
collector = _set_up_collector(module, job_id, queue, manager)
module.get_worker_timeout_seconds = lambda: 0.01
module.HEARTBEAT_INTERVAL = 0.1
module.probe_worker = lambda *_args, **_kwargs: pytest.fail("serverless slots are not worker URLs")
with pytest.raises(RuntimeError, match="serverless.*stopped.*result|serverless.*result.*stopped"):
asyncio.run(collector.execute(
images=None,
audio=None,
multi_job_id=job_id,
enabled_worker_ids='["serverless:pool:1"]',
delegate_only=True,
))
assert manager.owned_calls
assert manager.cancel_calls[0]["collector_job_id"] == job_id
assert job_id not in module.prompt_server.distributed_pending_jobs
def test_interruption_cancels_serverless_job_and_removes_collector_queue():
module = _load_collector_module()
job_id = "collector-interrupted"
queue = asyncio.Queue()
manager = _FakeManager(pending=True)
collector = _set_up_collector(module, job_id, queue, manager)
interrupted = module.comfy.model_management.InterruptProcessingException("stop")
module.comfy.model_management.throw_exception_if_processing_interrupted = lambda: (_ for _ in ()).throw(interrupted)
with pytest.raises(module.comfy.model_management.InterruptProcessingException, match="stop"):
asyncio.run(collector.execute(
images=None,
audio=None,
multi_job_id=job_id,
enabled_worker_ids='["serverless:pool:1"]',
delegate_only=True,
))
assert manager.cancel_calls[0]["collector_job_id"] == job_id
assert job_id not in module.prompt_server.distributed_pending_jobs
def test_serverless_tensor_shape_mismatch_raises_instead_of_silent_fallback():
module = _load_collector_module()
job_id = "collector-shape-mismatch"
queue = asyncio.Queue()
queue.put_nowait({
"worker_id": "serverless:pool:1",
"image_index": 0,
"tensor": torch.ones(1, 3, 2, 2),
"is_last": True,
})
manager = _FakeManager()
collector = _set_up_collector(module, job_id, queue, manager)
local_images = torch.zeros(1, 2, 2, 2)
with pytest.raises(RuntimeError, match="serverless.*shape|shape.*serverless"):
asyncio.run(collector.execute(
images=local_images,
audio=None,
multi_job_id=job_id,
enabled_worker_ids='["serverless:pool:1"]',
))
assert manager.cancel_calls[0]["collector_job_id"] == job_id
assert job_id not in module.prompt_server.distributed_pending_jobs
def test_regular_worker_timeout_keeps_partial_behavior_without_creating_manager():
module = _load_collector_module()
job_id = "collector-regular-timeout"
queue = asyncio.Queue()
collector = _set_up_collector(module, job_id, queue)
module.get_worker_timeout_seconds = lambda: 0.01
module.HEARTBEAT_INTERVAL = 0.1
local_images = torch.zeros(1, 2, 2, 2)
collector._get_serverless_manager = lambda *_args, **_kwargs: pytest.fail("regular collection must not create manager")
images, _audio = asyncio.run(collector.execute(
images=local_images,
audio=None,
multi_job_id=job_id,
enabled_worker_ids='["worker-a"]',
))
assert torch.equal(images, local_images)
assert job_id not in module.prompt_server.distributed_pending_jobs