| service: upload-processing |
| |
| runtime: python39 |
| # The default workers number is 1, increase it to improve throughput. |
| entrypoint: gunicorn -b:$PORT dashboard.dispatcher:APP --workers 8 --worker-class gthread --threads 10 --timeout 300 |
| app_engine_apis: true |
| # 3072MB Memory, 2.4GHz CPU limit, support auto scaling, support 8 workers. |
| instance_class: F4_1G |
| |
| automatic_scaling: |
| max_instances: 1800 |
| min_instances: 24 |
| # Define max_idle_instance to reduce the number of billed instance. |
| max_idle_instances: 30 |
| # 8 worker * 10 threads = 80 concurrent requests. |
| max_concurrent_requests: 80 |
| # Auto scaling will be triggered when reach 90% of the max concurrent. |
| target_throughput_utilization: 0.90 |
| target_cpu_utilization: 0.90 |
| max_pending_latency: automatic |
| |
| env_variables: |
| GAE_USE_SOCKETS_HTTPLIB: 'true' |
| |
| handlers: |
| - url: /.* |
| script: dashboard.dispatcher.APP |
| secure: always |