Feature wanted: 增加对每个访问目标URL的流量统计

[deanmax]

这会比较方便查看具体那些网站的流量占用比较多。目前只能看到total traffic。
YACD已经有类似功能，所以应该不太难实现

[Dreamacro]

这个只是单连接，clash dashboard 也有

[deanmax]

我的意思是，是否方便在metrics里加入url label:
比如 {"down":1434134,"type":"traffic","up":3724,"url":"www.google.com"}

[luckyyyyy]

@Dreamacro 作者你好 CLASH_TOKEN的值应该就是secret吧?

但是似乎我尝试直接访问
curl http://10.0.0.1:9090/traffic?token=123456 提示未授权 我看收集器就是这么连的ws 所以也失败了

换成文档写的方式就可以
curl -v -H "Authorization: Bearer 123456" http://10.0.0.1:9090/traffic

所以是?

[Dreamacro]

@luckyyyyy 只有 websocket 连接支持 query token，http 是不支持的

[Dreamacro]

@deanmax 你可以自己添加面板实现这个功能，仓库里只是个参考

[luckyyyyy]

@luckyyyyy 只有 websocket 连接支持 query token，http 是不支持的

原来如此，晚些我换ws再验证一下，我现在裸跑效果挺好的，链路收集都很正常，昨天试了很多次加上token都不行。

[deanmax]

@deanmax 你可以自己添加面板实现这个功能，仓库里只是个参考

添加面板简单，只要metrics有支持按url的流量统计即可(现在应该还没有可以用的metrics)

[Xm798]

@deanmax 你可以自己添加面板实现这个功能，仓库里只是个参考

请问有按节点统计流量的 metrics 吗？

[khronosmoe]

I don't think there is a metric that supports exporting statistical information about a specific site/node/interface/user. What I get from ws://$CLASH_HOST/traffic?token=$CLASH_TOKEN is only the current upload and download speed.
Unlike V2ray's stats api which can output the statistical usage of each interface:

name: "inbound>>>ss>>>traffic>>>downlink"
value: 1704
name: "inbound>>>ss>>>traffic>>>uplink"
value: 2276
name: "user>>>u9@ss>>>traffic>>>uplink"
value: 1776
name: "user>>>u9@ss>>>traffic>>>downlink"
value: 1368
name: "user>>>u3@ws>>>traffic>>>uplink"
value: 2810759
name: "user>>>u3@ws>>>traffic>>>downlink"
value: 126944108

What I get when accessing ws://$CLASH_HOST/profile/tracing?token=$CLASH_TOKEN:

{
    "answer": "xx.xx.xx.xx",
    "dnsType": "fakeip",
    "duration": 5,
    "id": "9a9axxxxxxxxxxxxxxx9a9a",
    "name": "www.apple.com",
    "qType": "AAAA",
    "type": "DNSRequest"
},
{
    "duration": 50000,
    "id": "4b4bxxxxxxxxxxxxxxx4b4b",
    "metadata": {
        "network": "tcp",
        "type": "TUN",
        "sourceIP": "198.18.0.1",
        "destinationIP": "xx.xx.xx.xx",
        "sourcePort": "40001",
        "destinationPort": "443",
        "host": "www.apple.com",
        "dnsMode": "fake-ip",
        "processPath": "ProcessName"
    },
    "payload": "",
    "proxy": "PROXY",
    "rule": "Match",
    "type": "RuleMatch"
},
{
    "address": "www.apple.com:443",
    "chain": [
        "NodeName",
        "PROXY"
    ],
    "duration": 150000,
    "host": "www.apple.com",
    "id": "4b4bxxxxxxxxxxxxxxx4b4b",
    "proxy": "NodeName",
    "type": "ProxyDial"
}

I can already get most of the metrics, with the exception of dial-up time, from clash.log and import them into Grafana using Promtail. The data from clash.log:

time="2022-07-22T00:00:00" level=debug msg="[Process] www.apple.com from process /ProcessName"
time="2022-07-22T00:00:00" level=info msg="[TCP] 198.18.0.1:40001 --> www.apple.com:443 match RuleSet(my-rule-set) using PROXY[NodeName]"