Python应用Datadog中冗余Gunicorn日志无法过滤的解决问询

问题 #

我的Python服务每次运行时都会生成以下日志：

"[2024-05-13 05:58:21 +0000] [67] [INFO] Booting worker with pid: 67"
"[2024-05-13 05:58:21 +0000] [1] [INFO] Using worker: sync"
"[2024-05-13 05:58:21 +0000] [1] [INFO] Listening at: http://0.0.0.0:5000,http://0.0.0.0:8080 (1)"
"[2024-05-13 05:58:21 +0000] [1] [INFO] Starting gunicorn 21.2.0"
"[2024-05-13 05:58:21 +0000] [67] [INFO] Worker exiting (pid: 67)"
"[2024-05-13 05:58:21 +0000] [1] [INFO] Shutting down: Master"
"[2024-05-13 05:58:21 +0000] [1] [INFO] Handling signal: term"

我尝试用以下自定义logging过滤器代码禁用这些日志，但并未生效：

import logging
import re

class CustomFilter(logging.Filter):
    def __init__(self, blocked_patterns):
        super().__init__()
        self.blocked_patterns = blocked_patterns

    def filter(self, record):
        # Check if the log message matches any of the blocked patterns
        for pattern in self.blocked_patterns:
            if re.match(pattern, record.getMessage()):
                # If a blocked pattern is found, return False to filter out the log record
                 return False
        # If no blocked pattern matches, allow the log record to pass through
        return True

# Configure logging
logging.basicConfig(level=logging.INFO)

# Define the list of patterns to be blocked
blocked_patterns = [
    r".*\[INFO\] Booting worker with pid: \d+",
    r".*\[INFO\] Using worker: .*",
    r".*\[INFO\] Listening at: http://0\.0\.0\.0:5000,http://0\.0\.0\.0:8080 \(1\)",
    r".*\[INFO\] Starting gunicorn \d+\.\d+\.\d+",
    r".*\[INFO\] Worker exiting \(pid: \d+\)",
    r".*\[INFO\] Shutting down: Master",
    r".*\[INFO\] Handling signal: term",
    r".*http://metadata\.google\.internal/computeMetadata/.*"
]

# Create a logger
logger = logging.getLogger()

# Add the custom filter to the logger
logger.addFilter(CustomFilter(blocked_patterns))

解决方案 #

核心原因 #

Gunicorn拥有独立的日志系统，它的日志输出不会经过你配置的根logger过滤器。Gunicorn会创建专属的logger实例（如gunicorn.error、gunicorn.info等），你的过滤器仅作用于根logger，对这些专用logger无效。

解决步骤 #

1. 给Gunicorn的logger添加过滤器
   直接获取Gunicorn相关的logger实例，为它们添加自定义过滤器：

   filter_instance = CustomFilter(blocked_patterns)
   # 遍历Gunicorn的所有相关logger并添加过滤器
   for logger_name in ['gunicorn', 'gunicorn.access', 'gunicorn.error', 'gunicorn.info']:
       gunicorn_logger = logging.getLogger(logger_name)
       gunicorn_logger.addFilter(filter_instance)
   

2. 简化正则匹配逻辑（可选）
   record.getMessage()返回的是原始日志消息内容，不包含时间戳、进程ID等前缀。比如实际日志消息是"Booting worker with pid: 67"，而非带时间戳的完整字符串。可以简化正则表达式，仅匹配消息部分：

   blocked_patterns = [
       r"Booting worker with pid: \d+",
       r"Using worker: .*",
       r"Listening at: http://0\.0\.0\.0:5000,http://0\.0\.0\.0:8080 \(1\)",
       r"Starting gunicorn \d+\.\d+\.\d+",
       r"Worker exiting \(pid: \d+\)",
       r"Shutting down: Master",
       r"Handling signal: term",
       r".*http://metadata\.google\.internal/computeMetadata/.*"
   ]
   

   这样匹配更精准，也能避免前缀格式变化导致的匹配失败。

3. 通过Gunicorn启动参数直接禁用（另一种思路）
   如果不需要Gunicorn的所有启动/关闭日志，可以在启动时通过命令行参数调整日志级别：

   gunicorn --log-level warning your_app:app
   

   将日志级别设为warning后，所有INFO级别的启动日志会被自动过滤。

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