fec656941d
CI/CD Pipeline / Dedup Check - skip PR tests when covered by push pipeline (pull_request) Successful in 6s
CI/CD Pipeline / Check if frontend-only change (pull_request) Successful in 4s
CI/CD Pipeline / PR Build API Image (pull_request) Successful in 24s
CI/CD Pipeline / PR Build Worker Image (pull_request) Successful in 24s
CI/CD Pipeline / Validate - Python (mypy + alembic) (pull_request) Successful in 1m41s
CI/CD Pipeline / Unit Tests (pull_request) Successful in 1m57s
CI/CD Pipeline / Integration Tests (pull_request) Successful in 1m57s
Preview Deploy / Deploy Preview Environment (pull_request) Successful in 1m44s
CI/CD Pipeline / Validate - Style (pull_request) Failing after 2m50s
CI/CD Pipeline / Validate - Security (pull_request) Has been cancelled
CI/CD Pipeline / Build Production API Image (pull_request) Has been cancelled
CI/CD Pipeline / Build Production Web Image (pull_request) Has been cancelled
CI/CD Pipeline / Build Production Worker Image (pull_request) Has been cancelled
CI/CD Pipeline / Deploy Production (pull_request) Has been cancelled
CI/CD Pipeline / Production Browser E2E (pull_request) Has been cancelled
CI/CD Pipeline / Canary Release to Production (pull_request) Has been cancelled
CI/CD Pipeline / CI Gate (pull_request) Has been cancelled
AI Code Review / AI Code Review (pull_request) Has been cancelled
PR Automation / Auto Approve on CI Green (pull_request) Has been cancelled
PR Automation / Auto Merge on CI Green + Approved (pull_request) Has been cancelled
CI/CD Pipeline / ACR Image Cleanup (pull_request) Failing after 241h52m35s
CI/CD Pipeline / Staging E2E Tests (pull_request) Failing after 241h52m38s
CI/CD Pipeline / Deploy Staging (Watchtower auto-deploy) (pull_request) Failing after 241h52m43s
CI/CD Pipeline / Retag skipped Staging Web Image (pull_request) Failing after 241h52m49s
CI/CD Pipeline / Retag skipped Staging API Image (pull_request) Failing after 241h52m49s
CI/CD Pipeline / Build Staging API Image (pull_request) Failing after 241h53m9s
CI/CD Pipeline / Build Staging Worker Image (pull_request) Failing after 241h53m9s
CI/CD Pipeline / PR Build Web Image (pull_request) Failing after 241h53m15s
CI/CD Pipeline / Check push changed paths (pull_request) Failing after 241h53m26s
CI/CD Pipeline / Frontend Unit Tests (pull_request) Failing after 241h53m16s
CI/CD Pipeline / Staging API Integration Tests (pull_request) Failing after 242h27m12s
CI/CD Pipeline / Retag skipped Staging Worker Image (pull_request) Failing after 242h27m25s
CI/CD Pipeline / Build Staging Web Image (pull_request) Failing after 242h27m45s
CI/CD Pipeline / Frontend Lint (pull_request) Failing after 242h27m52s
根因(staging 实证 IMG_2285.MOV file_size=0 卡 processing 持续误杀):
direct complete 不传 file_size,find_recent_active_by_library_and_name
的大小校验 if file_size>0 不生效,30 分钟内同名视频(iPhone IMG_xxxx.MOV)
即使内容全新也被同名兜底误判重复跳过。
兜底去重收紧(宁可漏判不可误杀):
- _find_duplicate_asset:file_hash/client_upload_id 非空时不走同名兜底
(hash 已代表内容);走到兜底必须 file_size>0 且与记录大小严格一致
- find_recent_active_by_library_and_name:file_size=0 直接返回 None,
大小条件改为 SQL 内严格等值匹配
- complete 的 _create_pending_asset 补传 file_size(之前占位记录大小永远 0)
ingest 链路孤儿兜底(此前只有 generation 链路有清理):
- 新增 packages/application/ingest_orphan_cleanup.py:
- processing>60min / pending>90min 的 ingest_job 标 failed,关联
processing/uploading asset 联动标 error;无 job 关联超 120min 孤儿
占位 asset 也标 error(beat 每 10 分钟巡检)
- worker 启动恢复:processing 超 10 分钟的 job CAS 重置 pending 并
重新派单(Redis SET NX 锁互斥双 worker,旧消息重投由执行前守卫丢弃)
- celery 配置:task_reject_on_worker_lost=True;
broker visibility_timeout=4h(acks_late 下长转码任务不被误重投)
单测:同名不同大小放行 / hash 非空不走兜底 / file_size=0 放行 /
同名同大小 processing 才判重;孤儿清理 9 例、启动恢复 4 例、
仓储严格守卫 6 例、beat 串联 2 例。diff coverage 100%。
前端配合(前端工程师):completeDirectUpload 请求体补 file_size=file.size。
65 lines
3.0 KiB
Python
Executable File
65 lines
3.0 KiB
Python
Executable File
from celery import Celery
|
||
from worker_app.core.config import get_settings
|
||
|
||
settings = get_settings()
|
||
celery_app = Celery(settings.worker_name)
|
||
celery_app.conf.broker_url = settings.broker_url
|
||
celery_app.conf.result_backend = settings.result_backend
|
||
celery_app.conf.broker_connection_retry_on_startup = True
|
||
|
||
# #1714 队列隔离:generation(高优,独占 worker)/ transcode(素材转码)/ celery(默认)
|
||
from packages.shared.celery_queues import ( # noqa: E402
|
||
GENERATION_WORKER_PREFETCH_MULTIPLIER,
|
||
apply_queue_settings,
|
||
)
|
||
|
||
apply_queue_settings(celery_app)
|
||
# 长渲染任务预取 1,避免任务被预取占住导致调度不均
|
||
celery_app.conf.worker_prefetch_multiplier = GENERATION_WORKER_PREFETCH_MULTIPLIER
|
||
celery_app.conf.task_acks_late = True # worker 崩溃时未完成任务重回队列,由执行前守卫丢弃作废消息
|
||
# worker 进程被 OOM/容器硬杀时拒绝 ack,消息留在队列由其他 worker 接手
|
||
celery_app.conf.task_reject_on_worker_lost = True
|
||
# Redis broker 消息可见性超时(#1714):acks_late 下,消息被预取后 visibility_timeout
|
||
# 内未 ack 才会重投。长任务(ingest HEVC 转码 20-30 分钟、生成硬超时 11 分钟)
|
||
# 必须远大于最长执行时间,否则正常任务会在执行中被误重投;4 小时覆盖最长转码 + 余量。
|
||
celery_app.conf.broker_transport_options = {"visibility_timeout": 4 * 60 * 60}
|
||
|
||
celery_app.conf.imports = (
|
||
"worker_app.tasks.health",
|
||
"worker_app.tasks.ingest",
|
||
"worker_app.tasks.classification",
|
||
"worker_app.tasks.generation",
|
||
"worker_app.tasks.voice_extraction",
|
||
"worker_app.tasks.voice_clone",
|
||
"worker_app.tasks.tts_synthesis",
|
||
"worker_app.tasks.batch_download",
|
||
"worker_app.tasks.duplication_check",
|
||
"worker_app.tasks._startup",
|
||
"apps.worker.video_processing.dedup",
|
||
"worker_app.tasks.cleanup",
|
||
)
|
||
|
||
# Celery Beat 定时任务调度
|
||
# 注:worker 单实例内嵌 beat(entrypoint-worker.sh -B),定时任务不会重复执行
|
||
celery_app.conf.beat_schedule = {
|
||
# pending 任务超时清理:worker 停止消费后,卡 pending 的任务 15 分钟内释放限流名额
|
||
"cleanup-stale-pending-tasks": {
|
||
"task": "worker.cleanup_stale_pending_tasks",
|
||
"schedule": 300.0, # 每 5 分钟(秒)
|
||
"options": {"expires": 240}, # 4 分钟过期,避免堆积
|
||
},
|
||
# running 孤儿任务巡检:容器重启/进程被杀后卡 running 的任务,20 分钟无更新则判失败
|
||
"cleanup-stale-running-tasks": {
|
||
"task": "worker.cleanup_stale_running_tasks",
|
||
"schedule": 300.0, # 每 5 分钟(秒)
|
||
"options": {"expires": 240},
|
||
},
|
||
# 上传/转码链路孤儿巡检:worker 重启丢 prefetch 消息后,卡 pending/processing
|
||
# 的 ingest_job + asset 占位超时标终态(#1714)。转码任务较长,10 分钟一轮
|
||
"cleanup-stale-ingest-jobs": {
|
||
"task": "worker.cleanup_stale_ingest_jobs",
|
||
"schedule": 600.0, # 每 10 分钟(秒)
|
||
"options": {"expires": 540},
|
||
},
|
||
}
|