"""Gunicorn config""" import os import pwd from multiprocessing import cpu_count import structlog from kubernetes.config.incluster_config import SERVICE_HOST_ENV_NAME bind = "127.0.0.1:8000" reload = True try: pwd.getpwnam("authentik") user = "authentik" group = "authentik" except KeyError: pass worker_class = "uvicorn.workers.UvicornWorker" # Docker containers don't have /tmp as tmpfs if os.path.exists("/dev/shm"): # nosec worker_tmp_dir = "/dev/shm" # nosec os.environ.setdefault("DJANGO_SETTINGS_MODULE", "authentik.root.settings") logconfig_dict = { "version": 1, "disable_existing_loggers": False, "formatters": { "json_formatter": { "()": structlog.stdlib.ProcessorFormatter, "processor": structlog.processors.JSONRenderer(), "foreign_pre_chain": [ structlog.stdlib.add_log_level, structlog.stdlib.add_logger_name, structlog.processors.TimeStamper(), structlog.processors.StackInfoRenderer(), ], } }, "handlers": { "error_console": { "class": "logging.StreamHandler", "formatter": "json_formatter", }, "console": {"class": "logging.StreamHandler", "formatter": "json_formatter"}, }, "loggers": { "uvicorn": {"handlers": ["console"], "level": "WARNING", "propagate": False}, "gunicorn": {"handlers": ["console"], "level": "INFO", "propagate": False}, }, } # if we're running in kubernetes, use fixed workers because we can scale with more pods # otherwise (assume docker-compose), use as much as we can if SERVICE_HOST_ENV_NAME in os.environ: default_workers = 2 else: default_workers = max(cpu_count() * 0.25, 1) + 1 # Minimum of 2 workers workers = int(os.environ.get("WORKERS", default_workers)) threads = int(os.environ.get("THREADS", 4))