录制报错，IO跑满

[echome123]

最近几天使用了V4.4.14版本的集成包录制，以前的旧版本稳定性基本没问题，但是最近突然频繁出现服务器宕机，查看了一下监控，磁盘IO突然跑满，并且日志报错 i/o timeout ：

正常情况下，磁盘IO很低，最近版本偶尔突然会出现IO持续跑满，导致服务器无法连接，API接口超时无响应，只能强制重启服务器才会恢复，相关日志如下：
`
11:03:15 �[31mERROR�[0m pull {"plugin": "HDL", "url": "http://xxxx.flv", "error": "read tcp 170.31.103.114:59754->60.221.209.176:80: use of closed network connection"}

11:17:55 �[31mERROR�[0m pull {"plugin": "HDL", "url": "xxx.flv", "error": "EOF"}

11:36:35 �[31mERROR�[0m pull {"plugin": "HDL", "url": "xxx.flv", "error": "read tcp 170.31.103.114:38960->1.1.124.161:80: read: connection reset by peer"}

11:36:37 �[31mERROR�[0m pull {"plugin": "HDL", "url": "http://xxx.flv", "error": "read tcp 170.31.103.114:44296->1.1.193.214:80: read: connection reset by peer"}

11:40:35 �[31mERROR�[0m pull connect {"plugin": "HDL", "url": "xxx.flv", "error": "Get "http:/xxx.flv": dial tcp: lookup pull-l.xxx.com: i/o timeout"}

//从这里往后就开始出故障了

11:40:34 �[33mWARN�[0m track timeout {"stream": "1/37-20230314095641", "name": "h264", "lastWriteTime": "11:36:37", "timeout": "10s"}
11:40:37 �[34mINFO�[0m 🟢�[33m->�[0m⌛ {"stream": "1/37-20230314095641", "action": "publish lost"}
11:40:37 �[33mWARN�[0m track timeout {"stream": "1/43-20230314111111", "name": "aac", "lastWriteTime": "11:36:35", "timeout": "10s"}
11:40:37 �[34mINFO�[0m 🟢�[33m->�[0m⌛ {"stream": "1/43-20230314111111", "action": "publish lost"}
11:40:37 �[34mINFO�[0m ⌛�[33m->�[0m🔴 {"stream": "1/37-20230314095641", "action": "timeout"}

11:40:37 �[31mERROR�[0m pull connect {"plugin": "HDL", "url": "http://xxx.flv", "error": "Get "http://xxx.flv": dial tcp: lookup pull-l.xxxxx.com: i/o timeout"}
11:40:37 �[34mINFO�[0m 🟢�[33m->�[0m⌛ {"stream": "1/39-20230314110933", "action": "publish lost"}
11:40:35 �[34mINFO�[0m pull {"plugin": "HDL", "path": "1/38-20230314113830", "url": "http://xxx.flv"}
11:40:37 �[34mINFO�[0m ⌛�[33m->�[0m🔴 {"stream": "1/43-20230314111111", "action": "timeout"}

//再往后就没有日志记录了，磁盘IO就降不下来了
`

[langhuihui]

升级到4.4.16

[echome123]

更新之后，刚才又出现了这样的情况

[langhuihui]

配置文件让我看看

[echome123]

global:
  loglevel: info # 日志级别，可选值：debug,info,warn,error,panic,fatal
  http:
    # 网关地址，用于访问API
    listenaddr: :8186
    # 用户名和密码，用于API访问时的基本身份认证
    username: "test"
    password: "123456"
hook:
  urllist:
    "*": "http://live.xxx.com"
record:
  flv:
    autorecord: true
    fragment: 18000s

其他配置项均为默认配置，只手动修改了上述配置，每次故障前后的m7s日志上面已经贴出，刚刚看到新版本在m7s根目录下还有一个 fatal.log的日志，内容如下：

--------------------------------

--------------------------------

--------------------------------
2023/03/15 04:30:29 http: URL query contains semicolon, which is no longer a supported separator; parts of the query may be stripped when parsed; see golang.org/issue/25192

[langhuihui]

你的流是从哪儿来的呢

[echome123]

网上搜集的一些源， 那我再尝试换其他的源观察观察

[langhuihui]

我看你的日志是从远端拉的flv流再录制到本地是吗？

[echome123]

我看你的日志是从远端拉的flv流再录制到本地是吗？

是的，都是远程拉的FLV，录制到本地记录