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.
This commit is contained in:
Robert Wojciechowski
+65
-14
@@ -35,7 +35,9 @@ from .orchestration.prompt_transform import (
|
||||
prepare_delegate_master_prompt,
|
||||
prune_prompt_for_worker,
|
||||
WORKER_SINK_CLASS,
|
||||
_resolve_direct_workers_for_image_list,
|
||||
)
|
||||
from ..image_list.runtime import validate_image_list_regions
|
||||
|
||||
|
||||
prompt_server = server.PromptServer.instance
|
||||
@@ -252,6 +254,8 @@ async def orchestrate_distributed_execution(
|
||||
execution_trace_id = trace_execution_id or _generate_execution_trace_id()
|
||||
|
||||
config = load_config()
|
||||
if delegate_master is None:
|
||||
delegate_master = bool(config.get("settings", {}).get("master_delegate_only", False))
|
||||
use_websocket = bool(config.get("settings", {}).get("websocket_orchestration", False))
|
||||
master_url = build_master_url(config=config, prompt_server_instance=prompt_server)
|
||||
(
|
||||
@@ -261,8 +265,42 @@ async def orchestrate_distributed_execution(
|
||||
media_sync_timeout_seconds,
|
||||
) = _resolve_orchestration_limits(config)
|
||||
requested_ids = enabled_worker_ids if enabled_worker_ids is not None else None
|
||||
workers = _resolve_enabled_workers(config, requested_ids)
|
||||
prompt_index = PromptIndex(prompt_obj)
|
||||
image_list_start_ids = prompt_index.nodes_for_class("DistributedImageListStart")
|
||||
image_list_mode = bool(image_list_start_ids)
|
||||
if image_list_mode:
|
||||
# Validate before probing or queueing anything so malformed boundaries and
|
||||
# old worker releases fail before any upstream image is materialized.
|
||||
validate_image_list_regions(prompt_obj)
|
||||
other_distributed = (
|
||||
prompt_index.nodes_for_class("DistributedCollector")
|
||||
+ prompt_index.nodes_for_class("UltimateSDUpscaleDistributed")
|
||||
)
|
||||
if other_distributed:
|
||||
raise ValueError(
|
||||
"Distributed Image List v1 cannot be mixed with Distributed Collector or "
|
||||
"Ultimate SD Upscale nodes in one prompt"
|
||||
)
|
||||
configured_serverless = [
|
||||
str(worker.get("id")) for worker in config.get("workers", [])
|
||||
if worker.get("enabled", False)
|
||||
and str(worker.get("type") or "local").lower() == "serverless"
|
||||
and (requested_ids is None or str(worker.get("id")) in {str(x) for x in requested_ids})
|
||||
]
|
||||
if configured_serverless:
|
||||
trace_debug(
|
||||
execution_trace_id,
|
||||
"Image List v1 explicitly excludes unsupported serverless pools: "
|
||||
+ ", ".join(configured_serverless),
|
||||
)
|
||||
workers = _resolve_direct_workers_for_image_list(config, requested_ids)
|
||||
if not workers and delegate_master:
|
||||
raise RuntimeError(
|
||||
"Distributed Image List in orchestrator-only mode requires a selected direct worker; "
|
||||
"serverless pools are unsupported"
|
||||
)
|
||||
else:
|
||||
workers = _resolve_enabled_workers(config, requested_ids)
|
||||
load_balance_requested = _prompt_requests_load_balance(prompt_index)
|
||||
trace_debug(
|
||||
execution_trace_id,
|
||||
@@ -278,26 +316,37 @@ async def orchestrate_distributed_execution(
|
||||
),
|
||||
)
|
||||
|
||||
# Respect master delegate-only configuration
|
||||
if delegate_master is None:
|
||||
delegate_master = bool(config.get("settings", {}).get("master_delegate_only", False))
|
||||
|
||||
if not workers and delegate_master:
|
||||
# Keep the legacy Collector fallback, but never override image-list
|
||||
# orchestrator-only mode merely because its direct workers are unavailable.
|
||||
if not image_list_mode and not workers and delegate_master:
|
||||
trace_debug(
|
||||
execution_trace_id,
|
||||
"Delegate-only requested but no workers are enabled. Falling back to master execution.",
|
||||
)
|
||||
delegate_master = False
|
||||
|
||||
active_workers, delegate_master = await select_active_workers(
|
||||
active_workers, selected_delegate_master = await select_active_workers(
|
||||
workers,
|
||||
use_websocket,
|
||||
delegate_master,
|
||||
False if image_list_mode else delegate_master,
|
||||
trace_execution_id=execution_trace_id,
|
||||
probe_concurrency=worker_probe_concurrency,
|
||||
)
|
||||
if image_list_mode:
|
||||
from .orchestration.dispatch import preflight_image_list_workers
|
||||
if not active_workers and delegate_master:
|
||||
raise RuntimeError(
|
||||
"No selected direct worker is reachable for Distributed Image List in orchestrator-only mode"
|
||||
)
|
||||
idle_ids = (await preflight_image_list_workers(active_workers, require_idle=delegate_master)
|
||||
if active_workers else {})
|
||||
active_workers = [worker for worker in active_workers if worker["id"] in idle_ids]
|
||||
if not active_workers and delegate_master:
|
||||
raise RuntimeError("No feature-compatible direct worker is idle in orchestrator-only mode")
|
||||
else:
|
||||
delegate_master = selected_delegate_master
|
||||
|
||||
if load_balance_requested:
|
||||
if load_balance_requested and not image_list_mode:
|
||||
candidate_workers = list(active_workers)
|
||||
if not delegate_master:
|
||||
# Include master in load balancing only when master participation is enabled.
|
||||
@@ -377,9 +426,10 @@ async def orchestrate_distributed_execution(
|
||||
master_url,
|
||||
delegate_master,
|
||||
prompt_index,
|
||||
workflow_meta=workflow_meta,
|
||||
)
|
||||
|
||||
if delegate_master:
|
||||
if delegate_master and not image_list_mode:
|
||||
collector_ids = find_nodes_by_class(master_prompt, "DistributedCollector")
|
||||
upscale_nodes = find_nodes_by_class(master_prompt, "UltimateSDUpscaleDistributed")
|
||||
if upscale_nodes:
|
||||
@@ -400,7 +450,7 @@ async def orchestrate_distributed_execution(
|
||||
+ ", ".join(f"{worker['name']} ({worker['id']})" for worker in active_workers),
|
||||
)
|
||||
worker_payloads = []
|
||||
if active_workers:
|
||||
if active_workers and not image_list_mode:
|
||||
worker_prep_semaphore = asyncio.Semaphore(worker_prep_concurrency)
|
||||
media_sync_semaphore = asyncio.Semaphore(media_sync_concurrency)
|
||||
worker_payloads = await asyncio.gather(
|
||||
@@ -431,8 +481,9 @@ async def orchestrate_distributed_execution(
|
||||
manager.ensure_capacity(len(serverless_jobs))
|
||||
|
||||
# Do not leave collector queues behind when offline compilation/media validation fails.
|
||||
for job_id in job_id_map.values():
|
||||
await _ensure_distributed_queue(job_id)
|
||||
if not image_list_mode:
|
||||
for job_id in job_id_map.values():
|
||||
await _ensure_distributed_queue(job_id)
|
||||
|
||||
if regular_payloads:
|
||||
await asyncio.gather(
|
||||
@@ -483,4 +534,4 @@ async def orchestrate_distributed_execution(
|
||||
execution_trace_id,
|
||||
f"Orchestration complete: prompt_id={prompt_id}, dispatched_workers={len(worker_payloads)}, delegate_master={delegate_master}",
|
||||
)
|
||||
return prompt_id, prompt_number, len(worker_payloads), node_errors
|
||||
return prompt_id, prompt_number, (len(active_workers) if image_list_mode else len(worker_payloads)), node_errors
|
||||
|
||||
Reference in New Issue
Block a user