Ubuntu服务器Flask应用报FileNotFoundError，Windows运行正常

问题描述 #

在DigitalOcean Ubuntu Droplet上部署Flask应用时返回内部服务器错误，错误提示找不到路径./images/ISIC_0029612.jpg.jpg，但当前代码中无任何涉及该路径的读写逻辑。本地Windows 11环境运行正常，已尝试重启/重载uwsgi、nginx服务，问题未解决。

错误堆栈信息 #

File "/var/www/html/cancel-the-cancer/venv/lib/python3.10/site-packages/flask/views.py", line 107, in view
    return current_app.ensure_sync(self.dispatch_request)(**kwargs)
  File "/var/www/html/cancel-the-cancer/venv/lib/python3.10/site-packages/flask_restful/__init__.py", line 582, in dispatch_request
    resp = meth(*args, **kwargs)
  File "/var/www/html/cancel-the-cancer/Resources/predict.py", line 10, in post
    img = img.resize((256, 256))
  File "/var/www/html/cancel-the-cancer/venv/lib/python3.10/site-packages/werkzeug/datastructures.py", line 3002, in save
    dst = open(dst, "wb")
FileNotFoundError: [Errno 2] No such file or directory: './images/ISIC_0029612.jpg.jpg'

相关代码 #

predict.py #

import numpy as np
from flask_restful import Resource, request
import tensorflow as tf
from PIL import Image

class Predict(Resource):
    def post(self):
        file = request.files['image']
        img = Image.open(file.stream)
        img = img.resize((256, 256))
        img = np.array(img)
        img = img / 255.0
        img = np.expand_dims(img, axis=0)
        loaded_model = tf.keras.models.load_model('../model/vgg19_92%_93%')
        prediction = loaded_model.predict(img)[0][0]
        return {"result": float(prediction)}

app.py #

import flask_restful
from flask_cors import CORS

from flask_cors import CORS
from flask import Flask
from flask_restful import Api

from Resources.predict import Predict

app = Flask(__name__)
app.secret_key = 'zwolnienizteorii44893'
api = Api(app)
cors = CORS(app)

api.add_resource(Predict, '/predict')

if __name__ == '__main__':
    app.run(port=5000, debug=True)

解决方案 #

1. 创建缺失的images目录并设置权限 #

虽然代码中未显式使用该目录，但依赖库可能隐式需要它，执行以下命令：

sudo mkdir /var/www/html/cancel-the-cancer/images
sudo chown www-data:www-data /var/www/html/cancel-the-cancer/images
sudo chmod 755 /var/www/html/cancel-the-cancer/images

2. 统一PIL/Pillow版本（本地与服务器一致） #

不同操作系统下的版本差异可能导致对象行为异常：

• 本地Windows执行pip show pillow获取版本号
• 服务器执行以下命令安装对应版本：

/var/www/html/cancel-the-cancer/venv/bin/pip install pillow==<本地版本号>

3. 修改代码确保Image对象完全加载到内存 #

避免流对象隐式触发磁盘操作，修改predict.py中图片处理部分：

file = request.files['image']
# 转换为RGB格式并确保加载到内存
img = Image.open(file.stream).convert('RGB')
img = img.resize((256, 256))

4. 检查uwsgi工作目录配置 #

确保uwsgi配置文件（如uwsgi.ini）中的chdir参数指向应用根目录：

chdir = /var/www/html/cancel-the-cancer

修改后重启uwsgi：

sudo systemctl restart uwsgi

内容的提问来源于stack exchange，提问作者Wiktor Fajkowski