[BUG] When running Pixiu via Docker, Pixiu fails to connect to the server on the host machine.

[Alanxtl]

Reproduction Process

Using dubbo-go-pixiu-samples as the service code, follow the HOWTO_CN.md in dubbo-go-pixiu-samples to reproduce the http/simple example.

You need to modify http/simple/pixiu/conf.yaml to observe the HTTP logs — add a new filter after line 53:

                  - name: dgp.filter.http.accesslog
                    config:
                      outPutPath: "/app/pixiu/access.log"

At this point, when running Pixiu, the server, and the client all locally, the example works properly (although the test may fail, the HTTP requests can still reach the server correctly).

The log from /app/pixiu/access.log looks like this:

[2025-08-04 15:07:17] 127.0.0.1:59150 -> localhost:8888 - cost time [ 26584 ] response [ {"message":"data is exist"}] 
[2025-08-04 15:07:17] 127.0.0.1:59150 -> localhost:8888 - cost time [ 23240 ] response [ {"id":"0001","name":"tc","age":18,"time":"2021-08-01T10:08:41Z"}]

Next, Pixiu is no longer run locally but instead run in Docker:

docker build --no-cache -t dubbo-go-pixiu:local .
docker run --name pixiu-gateway -p 8888:8888 -d -v /abs/path/to/dubbo-go-pixiu-samples/http/simple/pixiu/conf.yaml:/etc/pixiu/conf.yaml dubbo-go-pixiu:local

Pixiu starts successfully, but the health check continually fails:

2025-08-04T08:56:33.054Z INFO healthcheck/utils.go:39 [health check] health check for address "127.0.0.1:1314" failed, dial tcp 127.0.0.1:1314: i/o timeout

Then rerun the tests from the samples:

go test -v ./http/simple/test/ --count=1

At this point, the log in /app/pixiu/access.log inside Docker shows:

[2025-08-04 09:57:48] 172.17.0.1:55738 -> 127.0.0.1:8888 - cost time [ 9223372036854775807 ]invoke err [ [123 34 109 101 115 115 97 103 101 34 58 34 99 108 117 115 116 101 114 32 110 111 116 32 102 111 117 110 100 32 101 110 100 112 111 105 110 116 34 125]]  response can not convert to string] 
[2025-08-04 09:57:48] 172.17.0.1:55738 -> 127.0.0.1:8888 - cost time [ 9223372036854775807 ]invoke err [ [123 34 109 101 115 115 97 103 101 34 58 34 99 108 117 115 116 101 114 32 110 111 116 32 102 111 117 110 100 32 101 110 100 112 111 105 110 116 34 125]]  response can not convert to string] 

From the logs, we can see that the client’s requests to Pixiu (running in Docker) are received successfully, but Pixiu’s requests to the server fail.

[Gucvii]

sirui跟进

[baerwang]

docker run --network host --name pixiu-gateway -p 8888:8888 -d -v /abs/path/to/dubbo-go-pixiu-samples/http/simple/pixiu/conf.yaml:/etc/pixiu/conf.yaml dubbo-go-pixiu:local

[nanjiek]

I have tested this scenario. Running Pixiu in Docker with --network=host and the backend server on Windows works fine. The root cause is that in WSL2 the server binds only to 127.0.0.1, which is not accessible from Docker.

To fix it, the server inside WSL2 must be reachable from outside WSL. One option is to bind to 0.0.0.0, but that may expose the port to the whole network, which is unsafe in some environments. A safer choice is to:

• Bind only to the WSL2 interface IP (e.g. 172.27.x.x), or
• Use Windows port forwarding to expose the port to host.docker.internal.

In practice, three setups work:

1. Server on Windows + Pixiu in Docker → use host.docker.internal:PORT.
2. Server also in Docker → run both in the same Docker network and use the container name.
3. Server in WSL2 → bind to the WSL IP (instead of only 127.0.0.1), or forward the port to Windows, then access via host.docker.internal:PORT.

---

我测试过这个场景：Pixiu 在 Docker 中用 --network=host 运行，后端服务在 Windows 上启动时没有问题。问题的根源是：在 WSL2 中如果服务只监听 127.0.0.1，Docker 容器是无法访问到的。

要解决，需要让 WSL2 中的服务能够被外部访问。虽然可以直接监听 0.0.0.0，但这样会对整个网络暴露端口，在某些环境下存在安全风险。更安全的做法是：

• 只绑定到 WSL2 的内部 IP（如 172.27.x.x），或者
• 在 Windows 上做端口转发，把端口暴露到 host.docker.internal。

实际可行的三种运行方式：

1. 服务在 Windows + Pixiu 在 Docker → 使用 host.docker.internal:PORT。
2. 服务也放到 Docker → 两个容器放到同一网络，用容器名互访。
3. 服务在 WSL2 → 绑定到 WSL 的 IP（不要只用 127.0.0.1），或者通过 Windows 转发端口，再用 host.docker.internal:PORT 访问。