[Bug] high CPU usage when using grpc with trojan/vmess/vless

[remedynotavailable]

Verify steps

• [X] 确保你使用的是本仓库最新的的 clash 或 clash Alpha 版本 Ensure you are using the latest version of Clash or Clash Premium from this repository.
• [ ] 如果你可以自己 debug 并解决的话，提交 PR 吧 Is this something you can debug and fix? Send a pull request! Bug fixes and documentation fixes are welcome.
• [X] 我已经在 Issue Tracker 中找过我要提出的问题 I have searched on the issue tracker for a related issue.
• [X] 我已经使用 Alpha 分支版本测试过，问题依旧存在 I have tested using the dev branch, and the issue still exists.
• [X] 我已经仔细看过 Documentation 并无法自行解决问题 I have read the documentation and was unable to solve the issue.
• [X] 这是 Clash 核心的问题，并非我所使用的 Clash 衍生版本（如 OpenClash、KoolClash 等）的特定问题 This is an issue of the Clash core per se, not to the derivatives of Clash, like OpenClash or KoolClash.

Clash version

1.14.3 to alpha

What OS are you seeing the problem on?

Linux

Clash config

proxies:
  - name: US+VMESS-gRPC (2023-05-04)
    server: us4.socifiles.com
    port: 443
    type: vmess
    uuid: cee1c580-e4d9-11ed-baaa-1239d0255272
    alterId: 0
    cipher: auto
    tls: true
    skip-cert-verify: true
    servername: us4.socifiles.com
    network: grpc
    grpc-opts:
      grpc-service-name: vmgrpc
    udp: true

Clash log

2023/04/29 19:01:20 protocol error: received DATA after END_STREAM
2023/04/29 19:01:19 protocol error: received DATA after END_STREAM
2023/04/29 19:01:17 protocol error: received DATA after END_STREAM
2023/04/29 19:01:17 protocol error: received DATA after END_STREAM
2023/04/29 19:00:13 protocol error: received DATA after END_STREAM
2023/04/29 19:00:03 protocol error: received DATA after END_STREAM
2023/04/29 18:59:48 protocol error: received DATA after END_STREAM
2023/04/29 18:59:45 protocol error: received DATA after END_STREAM
2023/04/29 18:03:45 protocol error: received DATA after END_STREAM
2023/04/29 18:03:45 protocol error: received DATA after END_STREAM
2023/04/29 18:02:50 protocol error: received DATA after END_STREAM
2023/04/29 18:00:14 protocol error: received DATA after END_STREAM
2023/04/29 17:49:42 protocol error: received DATA after END_STREAM
2023/04/29 17:48:11 protocol error: received DATA after END_STREAM
2023/04/29 17:42:13 protocol error: received DATA after END_STREAM
2023/04/29 17:42:10 protocol error: received DATA after END_STREAM
2023/04/29 17:34:42 protocol error: received DATA after END_STREAM
2023/04/29 17:30:44 protocol error: received DATA after END_STREAM
2023/04/29 17:30:42 protocol error: received DATA after END_STREAM
2023/04/29 17:30:11 protocol error: received DATA after END_STREAM
2023/04/29 17:27:18 protocol error: received DATA after END_STREAM
2023/04/29 17:27:16 protocol error: received DATA after END_STREAM
2023/04/29 17:27:14 protocol error: received DATA after END_STREAM
2023/04/29 17:27:14 protocol error: received DATA after END_STREAM
2023/04/29 17:26:04 protocol error: received DATA after END_STREAM
2023/04/29 17:25:13 protocol error: received DATA after END_STREAM
2023/04/29 17:15:08 protocol error: received DATA after END_STREAM
2023/04/29 17:13:05 protocol error: received DATA after END_STREAM
2023/04/29 17:09:36 protocol error: received DATA after END_STREAM
2023/04/29 17:05:11 protocol error: received DATA after END_STREAM
2023/04/29 17:05:11 protocol error: received DATA after END_STREAM
2023/04/29 17:05:10 protocol error: received DATA after END_STREAM
2023/04/29 17:03:39 protocol error: received DATA after END_STREAM
2023/04/29 17:00:09 protocol error: received DATA after END_STREAM
2023/04/29 16:58:38 protocol error: received DATA after END_STREAM
2023/04/29 16:32:48 protocol error: received DATA after END_STREAM
2023/04/29 16:31:44 protocol error: received DATA after END_STREAM
2023/04/29 16:28:05 protocol error: received DATA after END_STREAM
2023/04/29 16:24:39 protocol error: received DATA after END_STREAM
2023/04/29 16:22:08 protocol error: received DATA after END_STREAM
2023/04/29 16:10:12 protocol error: received DATA after END_STREAM
2023/04/29 16:02:46 protocol error: received DATA after END_STREAM
2023/04/29 15:46:39 protocol error: received DATA after END_STREAM
2023/04/29 15:45:04 protocol error: received DATA after END_STREAM
2023/04/29 15:44:24 protocol error: received DATA after END_STREAM
2023/04/29 15:43:30 protocol error: received DATA after END_STREAM
2023/04/29 15:43:17 protocol error: received DATA after END_STREAM

Description

When using grpc for any trojan/vmess/vless the CPU usage went berserk. It may not immediately pickep up but after a few minutes used for browsing the CPU stuck at 100%. I confirmed it with other proxies the problem is with grpc. Just tried with any free grpc trojan/vmess/vless found on Google. Even after changing the proxy to non grpc it wouldn't goes down. Checked the yacd dashboard there's no ongoing heavy load. It's on idle, dl and ul just around 5kbps. Only after restarting the clash process and removing the grpc proxies it would back to normal. Tried the stable version from last 2022 to alpha it still the same. Also tried boxformagisk(with clash core) for Android the issue persists.

Thanks.

Picrel

[remedynotavailable]

Apparently it's an old issue that was ignored. https://github.com/Dreamacro/clash/issues/1979#issue-1143190088 It's affecting all h2 protocol. Maybe kindly add the option initial_windows_size

[Skyxim]

@remedynotavailable setting log level is debug, and the continuous usage of CPU is too high. Visit http://${contrller-api}/debug/pprof/profile, wait and upload the downloaded file

[remedynotavailable]

@remedynotavailable setting log level is debug, and the continuous usage of CPU is too high. Visit http://${contrller-api}/debug/pprof/profile, wait and upload the downloaded file

~ $ curl "http://192.168.1.1:9090/version"
{"meta":true,"version":"v1.14.4"}
~ $ curl "http://192.168.1.1:9090/debug/pprof/profile"
404 page not found
~ $

Do I need to compile it with debug tags or something?

[Skyxim]

@remedynotavailable Need to launch with log-level: debug

[remedynotavailable]

Running it with debug for more 3 hours already without cpu spike. Only higher ram and CPU than average. I will wait more.

[remedynotavailable]

Only 1 core 100% used at the moment. I will post again when all cores maxed out. profile (2).log

[remedynotavailable]

profile.txt

[remedynotavailable]

331 was correct it's only happened when used as upstream DNS by clash. I tried using adguardhome with clash+trojan grpc as upstream. there seems no problem so far though the log still print protocol error: received DATA after END_STREAM. but cpu is normal.

the problem only occured with internal clash dns resolver in combination with grpc.

[lux5am]

@Skyxim can you please consider to add full grpc like sing-box. It can be included with compile time flag. I tried with sing-box it's basically the same result. High CPU and RAM after a while. compiled sing-box with with_grpc tag, the result is very stable. Tho, the performance slightly degraded. Idk maybe just placebo. Tried with speedtest just a bit lower. I tried to do it myself but stuck here and there. I'm new in golang. Thanks.