updates :)

This commit is contained in:
2026-06-20 17:07:48 +03:30
parent 24155ad6bf
commit 64886b2b10
17 changed files with 896 additions and 262 deletions
+299 -113
View File
@@ -1,137 +1,306 @@
"""Huey task definitions and shared batch state for the admin panel."""
"""Batch task execution via RQ (Redis Queue) — fully decoupled from FastAPI's event loop."""
from __future__ import annotations
import threading
import json
import os
import sqlite3
import time
from concurrent.futures import ThreadPoolExecutor, as_completed
import uuid
from typing import Any
from huey import MemoryHuey
import redis
from rq import Queue
huey = MemoryHuey("seed", immediate=False)
from db.connection import DB_PATH
# ── Shared state (all accessed under _lock) ───────────────────────────────────
_lock = threading.Lock()
REDIS_URL = os.environ.get("REDIS_URL", "redis://localhost:6379/0")
_current_batch_id: str | None = None
_current_result: Any = None # Huey Result — kept for revocation
# Redis key helpers
_K_CURRENT = "seed:current" # string: current batch_id (or empty)
_K_META = "seed:meta:{}" # JSON string: batch aggregate meta
_K_SLOTS = "seed:slots:{}" # hash: slot_key → JSON worker info
_K_STOP = "seed:stop:{}" # string: "1" if stop requested
_K_QUEUE = "seed:pending_queue" # list: JSON-encoded pending items
_batch_meta: dict[str, Any] = {} # batch_id → aggregate metadata
_worker_slots: dict[str, dict] = {} # f"{batch_id}:{slot:04d}" → worker info
_stop_events: dict[str, threading.Event] = {}
def _r() -> redis.Redis: # type: ignore[type-arg]
return redis.from_url(REDIS_URL, decode_responses=True)
def _q() -> Queue:
return Queue("batch", connection=redis.from_url(REDIS_URL))
# ── Sync DB helpers ────────────────────────────────────────────────────────────
def _db_insert_run(
scenario: str,
identifier: str,
success: bool,
failure_reason: str | None,
steps: list[dict[str, Any]],
total_ms: int,
) -> None:
conn = sqlite3.connect(DB_PATH)
try:
conn.execute(
"INSERT INTO runs (scenario, identifier, success, failure_reason, steps_json, total_ms)"
" VALUES (?,?,?,?,?,?)",
(scenario, identifier, int(success), failure_reason, json.dumps(steps), total_ms),
)
conn.commit()
finally:
conn.close()
def _db_insert_batch_run(
batch_id: str,
config_id: int | None,
config_name: str,
workers: int,
total_runs: int,
stagger_ms: int,
headless: bool,
) -> None:
conn = sqlite3.connect(DB_PATH)
try:
conn.execute(
"INSERT INTO batch_runs"
" (batch_id, config_id, config_name, workers, total_runs, stagger_ms, headless)"
" VALUES (?,?,?,?,?,?,?)",
(batch_id, config_id, config_name, workers, total_runs, stagger_ms, int(headless)),
)
conn.commit()
finally:
conn.close()
def _db_finish_batch_run(
batch_id: str,
succeeded: int,
failed: int,
stopped: bool,
error: str | None,
) -> None:
conn = sqlite3.connect(DB_PATH)
try:
conn.execute(
"UPDATE batch_runs"
" SET finished_at=strftime('%Y-%m-%dT%H:%M:%SZ','now'),"
" succeeded=?, failed=?, stopped=?, error=?"
" WHERE batch_id=?",
(succeeded, failed, int(stopped), error, batch_id),
)
conn.commit()
finally:
conn.close()
# ── Public helpers ─────────────────────────────────────────────────────────────
def get_current_batch_id() -> str | None:
with _lock:
return _current_batch_id
def get_queue() -> list[dict[str, Any]]:
r = _r()
raw = r.lrange(_K_QUEUE, 0, -1)
items = []
for s in raw:
try:
item = json.loads(s)
items.append({k: v for k, v in item.items() if k != "cfg"})
except Exception:
pass
return items
def set_current(batch_id: str, result: Any) -> None:
global _current_batch_id, _current_result
with _lock:
_current_batch_id = batch_id
_current_result = result
def enqueue(
cfg: dict[str, Any],
workers: int,
total_runs: int,
stagger_ms: int,
headless: bool,
) -> dict[str, Any]:
queue_id = uuid.uuid4().hex
item: dict[str, Any] = {
"queue_id": queue_id,
"cfg": cfg,
"config_id": cfg.get("id"),
"config_name": cfg.get("name") or "",
"workers": workers,
"total_runs": total_runs,
"stagger_ms": stagger_ms,
"headless": headless,
"added_at": time.time(),
}
r = _r()
current = r.get(_K_CURRENT)
if current:
meta_raw = r.get(_K_META.format(current))
if meta_raw:
meta = json.loads(meta_raw)
if meta.get("running"):
r.rpush(_K_QUEUE, json.dumps(item))
return {"queue_id": queue_id, "status": "queued"}
_launch(item)
return {"queue_id": queue_id, "status": "started"}
def dequeue(queue_id: str) -> bool:
r = _r()
raw = r.lrange(_K_QUEUE, 0, -1)
for s in raw:
try:
item = json.loads(s)
if item.get("queue_id") == queue_id:
r.lrem(_K_QUEUE, 1, s)
return True
except Exception:
pass
return False
def clear_queue() -> int:
r = _r()
raw = r.lrange(_K_QUEUE, 0, -1)
count = len(raw)
r.delete(_K_QUEUE)
return count
def stop_current() -> bool:
"""Signal the running batch to stop. Returns True if there was something to stop."""
with _lock:
bid = _current_batch_id
result = _current_result
if bid is None:
r = _r()
current = r.get(_K_CURRENT)
if not current:
return False
# Revoke task if it hasn't started yet (still pending in Huey queue)
if result is not None:
try:
result.revoke(revoke_once=True)
except Exception:
pass
# Signal running workers to stop after their current step
ev = _stop_events.get(bid)
if ev:
ev.set()
with _lock:
if bid in _batch_meta:
_batch_meta[bid]["stopping"] = True
meta_raw = r.get(_K_META.format(current))
if not meta_raw:
return False
meta = json.loads(meta_raw)
if not meta.get("running"):
return False
r.set(_K_STOP.format(current), "1")
meta["stopping"] = True
r.set(_K_META.format(current), json.dumps(meta))
return True
def batch_status() -> dict[str, Any]:
with _lock:
bid = _current_batch_id
if bid is None:
return {"batch_id": None, "running": False}
r = _r()
bid = r.get(_K_CURRENT)
if not bid:
return {"batch_id": None, "running": False, "queue": get_queue()}
meta = dict(_batch_meta.get(bid, {}))
prefix = f"{bid}:"
workers = [dict(v) for k, v in _worker_slots.items() if k.startswith(prefix)]
meta_raw = r.get(_K_META.format(bid))
meta = json.loads(meta_raw) if meta_raw else {}
slot_data = r.hgetall(_K_SLOTS.format(bid))
workers = []
for v in slot_data.values():
try:
workers.append(json.loads(v))
except Exception:
pass
meta["batch_id"] = bid
meta["workers_detail"] = sorted(workers, key=lambda w: w.get("slot", 0))
meta["active_workers"] = sum(1 for w in workers if w.get("status") == "running")
meta["queue"] = get_queue()
return meta
# ── Huey task ─────────────────────────────────────────────────────────────────
# ── Internal ───────────────────────────────────────────────────────────────────
@huey.task(name="run_batch")
def run_batch_task(
batch_id: str,
cfg: dict[str, Any],
workers: int,
total_runs: int,
headless: bool,
stagger_ms: int,
) -> dict[str, Any]:
import asyncio
def _launch(item: dict[str, Any]) -> None:
batch_id = uuid.uuid4().hex
r = _r()
r.set(_K_CURRENT, batch_id)
q = _q()
q.enqueue(
run_batch_job,
batch_id,
item,
job_id=batch_id,
job_timeout=-1,
result_ttl=3600,
)
def _start_next_in_queue() -> None:
r = _r()
raw = r.lpop(_K_QUEUE)
if not raw:
return
try:
item = json.loads(raw)
except Exception:
return
_launch(item)
# ── RQ job (runs in worker process, no asyncio) ────────────────────────────────
def run_batch_job(batch_id: str, item: dict[str, Any]) -> None:
"""Top-level RQ job function. Runs in a forked worker process."""
import threading
from concurrent.futures import ThreadPoolExecutor, as_completed
from crawler.driver import driver_session
from crawler.dynamic_flow import DynamicFlow
from crawler.scenarios.fresh_session import FreshSessionScenario
from db.repositories.runs import runs_repo
from logger import get_logger
log = get_logger(__name__)
stop_ev = threading.Event()
with _lock:
_stop_events[batch_id] = stop_ev
_batch_meta[batch_id] = {
"running": True,
"stopping": False,
"total": total_runs,
"completed": 0,
"succeeded": 0,
"failed": 0,
"configured_workers": workers,
"started_at": time.time(),
"finished_at": None,
"error": None,
}
r = redis.from_url(REDIS_URL, decode_responses=True)
cfg = item["cfg"]
workers = item["workers"]
total_runs = item["total_runs"]
stagger_ms = item["stagger_ms"]
headless = item["headless"]
config_id: int | None = int(cfg["id"]) if cfg and cfg.get("id") is not None else None
config_name: str = str(cfg.get("name") or "") if cfg else ""
_db_insert_batch_run(batch_id, config_id, config_name, workers, total_runs, stagger_ms, headless)
meta: dict[str, Any] = {
"running": True,
"stopping": False,
"total": total_runs,
"completed": 0,
"succeeded": 0,
"failed": 0,
"configured_workers": workers,
"config_id": config_id,
"config_name": config_name,
"started_at": time.time(),
"finished_at": None,
"error": None,
}
r.set(_K_META.format(batch_id), json.dumps(meta))
slots_key = _K_SLOTS.format(batch_id)
stop_key = _K_STOP.format(batch_id)
meta_key = _K_META.format(batch_id)
_slots_lock = threading.Lock()
def _run_one(slot: int) -> bool:
slot_key = f"{batch_id}:{slot:04d}"
slot_key = f"{slot:04d}"
if stop_ev.is_set():
with _lock:
_worker_slots[slot_key] = {
"slot": slot, "status": "skipped",
"started_at": None, "finished_at": None, "duration_ms": None, "error": "stopped",
}
if r.exists(stop_key):
r.hset(slots_key, slot_key, json.dumps({
"slot": slot, "status": "skipped",
"started_at": None, "finished_at": None, "duration_ms": None, "error": "stopped",
}))
return False
with _lock:
_worker_slots[slot_key] = {
"slot": slot, "status": "running",
"started_at": time.time(), "finished_at": None, "duration_ms": None, "error": None,
}
r.hset(slots_key, slot_key, json.dumps({
"slot": slot, "status": "running",
"started_at": time.time(), "finished_at": None, "duration_ms": None, "error": None,
}))
t0 = time.monotonic()
identifier = f"batch-{batch_id[:8]}-{slot:04d}"
@@ -150,24 +319,23 @@ def run_batch_task(
{"name": s.name, "success": s.success, "error": s.error, "duration_ms": s.duration_ms}
for s in (result.steps if result else [])
]
asyncio.run(runs_repo.insert("batch", identifier, success, reason, steps, total_ms))
_db_insert_run("batch", identifier, success, reason, steps, total_ms)
status = "ok" if success else "failed"
with _lock:
_worker_slots[slot_key].update({
"status": status, "finished_at": time.time(), "duration_ms": total_ms,
})
r.hset(slots_key, slot_key, json.dumps({
"slot": slot, "status": status,
"finished_at": time.time(), "duration_ms": total_ms, "error": None,
}))
log.info("[batch:%s] slot %d %s%d ms", batch_id[:8], slot, status, total_ms)
return success
except Exception as exc:
total_ms = int((time.monotonic() - t0) * 1000)
asyncio.run(runs_repo.insert("batch", identifier, False, str(exc), [], total_ms))
with _lock:
_worker_slots[slot_key].update({
"status": "error", "error": str(exc),
"finished_at": time.time(), "duration_ms": total_ms,
})
_db_insert_run("batch", identifier, False, str(exc), [], total_ms)
r.hset(slots_key, slot_key, json.dumps({
"slot": slot, "status": "error", "error": str(exc),
"finished_at": time.time(), "duration_ms": total_ms,
}))
log.error("[batch:%s] slot %d error — %s", batch_id[:8], slot, exc)
return False
@@ -181,12 +349,17 @@ def run_batch_task(
) as pool:
futures = []
for i in range(total_runs):
if stop_ev.is_set():
if r.exists(stop_key):
break
futures.append(pool.submit(_run_one, i))
if stagger_ms > 0 and i < total_runs - 1:
if stop_ev.wait(timeout=stagger_ms / 1000):
break # stop was signalled during the stagger delay
deadline = time.monotonic() + stagger_ms / 1000
while time.monotonic() < deadline:
if r.exists(stop_key):
break
time.sleep(0.1)
if r.exists(stop_key):
break
for fut in as_completed(futures):
try:
@@ -197,18 +370,31 @@ def run_batch_task(
succeeded += 1
else:
failed += 1
with _lock:
_batch_meta[batch_id]["completed"] = succeeded + failed
_batch_meta[batch_id]["succeeded"] = succeeded
_batch_meta[batch_id]["failed"] = failed
current_meta_raw = r.get(meta_key)
if current_meta_raw:
current_meta = json.loads(current_meta_raw)
current_meta["completed"] = succeeded + failed
current_meta["succeeded"] = succeeded
current_meta["failed"] = failed
r.set(meta_key, json.dumps(current_meta))
except Exception as exc:
with _lock:
_batch_meta[batch_id]["error"] = str(exc)
current_meta_raw = r.get(meta_key)
if current_meta_raw:
current_meta = json.loads(current_meta_raw)
current_meta["error"] = str(exc)
r.set(meta_key, json.dumps(current_meta))
finally:
with _lock:
_batch_meta[batch_id]["running"] = False
_batch_meta[batch_id]["finished_at"] = time.time()
final_meta_raw = r.get(meta_key)
error_msg: str | None = None
if final_meta_raw:
final_meta = json.loads(final_meta_raw)
error_msg = final_meta.get("error")
final_meta["running"] = False
final_meta["finished_at"] = time.time()
r.set(meta_key, json.dumps(final_meta))
return {"batch_id": batch_id, "succeeded": succeeded, "failed": failed}
_db_finish_batch_run(batch_id, succeeded, failed, bool(r.exists(stop_key)), error_msg)
_start_next_in_queue()