gunicorn+uvicorn+fastapi 部署docker 内存占用过高问题

起因是我们的一个算法接口服务的镜像，在服务器上部署的时候，一下会占用60g的内存，但这个服务本身不到1g。业务需要对这个服务进行优化。 我逛遍了各种论坛也没有找到原因。 后来直接进入镜像查看配置文件。 gunicorn_conf.py

import json
import multiprocessing
import os

workers_per_core_str = os.getenv("WORKERS_PER_CORE", "1")
max_workers_str = os.getenv("MAX_WORKERS")
use_max_workers = None
if max_workers_str:
    use_max_workers = int(max_workers_str)
web_concurrency_str = os.getenv("WEB_CONCURRENCY", None)

host = os.getenv("HOST", "0.0.0.0")
port = os.getenv("PORT", "80")
bind_env = os.getenv("BIND", None)
use_loglevel = os.getenv("LOG_LEVEL", "info")
if bind_env:
    use_bind = bind_env
else:
    use_bind = f"{host}:{port}"

cores = multiprocessing.cpu_count()
workers_per_core = float(workers_per_core_str)
default_web_concurrency = workers_per_core * cores
if web_concurrency_str:
    web_concurrency = int(web_concurrency_str)
    assert web_concurrency > 0
else:
    web_concurrency = max(int(default_web_concurrency), 2)
    if use_max_workers:
        web_concurrency = min(web_concurrency, use_max_workers)
accesslog_var = os.getenv("ACCESS_LOG", "-")
use_accesslog = accesslog_var or None
errorlog_var = os.getenv("ERROR_LOG", "-")
use_errorlog = errorlog_var or None
graceful_timeout_str = os.getenv("GRACEFUL_TIMEOUT", "120")
timeout_str = os.getenv("TIMEOUT", "120")
keepalive_str = os.getenv("KEEP_ALIVE", "5")

# Gunicorn config variables
loglevel = use_loglevel
workers = web_concurrency
bind = use_bind
errorlog = use_errorlog
worker_tmp_dir = "/dev/shm"
accesslog = use_accesslog
graceful_timeout = int(graceful_timeout_str)
timeout = int(timeout_str)
keepalive = int(keepalive_str)


# For debugging and testing
log_data = {
    "loglevel": loglevel,
    "workers": workers,
    "bind": bind,
    "graceful_timeout": graceful_timeout,
    "timeout": timeout,
    "keepalive": keepalive,
    "errorlog": errorlog,
    "accesslog": accesslog,
    # Additional, non-gunicorn variables
    "workers_per_core": workers_per_core,
    "use_max_workers": use_max_workers,
    "host": host,
    "port": port,
}
print(json.dumps(log_data))

web_concurrency:如果在不设置的情况下，default_web_concurrency =min( workers_per_core * cores,use_max_workers)，也就是如果没有use_max_workers，那么服务器上有多少进程就启动多少个进程，一个进程等于一个完整的服务。 服务器上一共32核64进程，部署服务就会启64个服务。

通过修改default_web_concurrency或者use_max_workers就可以解决这个问题了。 解决方法是挺简单的，就是找不到，这里记录一下，万一有人和我一样遇到相同问题呢！