ERROR - Error in call_tool: Runtime service did not star 报错

[aGastronome]

沙箱服务在生产环境用docker容器部署，在启动后进行对话时，沙箱管理服务无法访问到容器中的工具。
具体日志如下：
INFO - Calling call_tool with data: {'identity': 'runtime_
sandbox_container_91fql6je294nwiuu231sw555r', 'tool_name': 'run_ipython_cell', 'arguments': {'code': '\n %reset -f -s\n
print("All variables and imports cleared")\n '}}
2026-02-02 02:45:13,529 - agentscope_runtime.sandbox.manager.server.app - ERROR - Error in call_tool: Runtime service did not start 。

[rayrayraykk]

Could you please refer to this tutorial and tell me more details?

[aGastronome]

容器内的日志如下：

root@3bbacb890183:/var/log# cat agentscope_runtime.err.log
INFO: Started server process [16]
INFO: Waiting for application startup.
Secure MCP Filesystem Server running on stdio
Allowed directories: [ '/workspace' ]
/agentscope_runtime/venv/lib/python3.11/site-packages/pydub/utils.py:170: RuntimeWarning: Couldn't find ffmpeg or avconv - defaultin
g to ffmpeg, but may not work
warn("Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work", RuntimeWarning)
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
INFO: Started server process [12]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
root@3bbacb890183:/var/log#

root@3bbacb890183:/var/log# cat nginx.err.log
2025/10/28 02:50:27 [warn] 19#19: duplicate extension "js", content type: "application/javascript", previous content type: "applicat
ion/javascript" in /etc/nginx/nginx.conf:30
2025/10/28 02:50:31 [error] 20#20: *1 connect() failed (111: Connection refused) while connecting to upstream, client: 172.17.0.1, s
erver: , request: "GET /fastapi/healthz HTTP/1.1", upstream: "http://127.0.0.1:8000/healthz", host: "localhost:8083"
2025/10/28 02:50:31 [error] 20#20: *1 connect() failed (111: Connection refused) while connecting to upstream, client: 172.17.0.1, s
erver: , request: "GET /fastapi/healthz HTTP/1.1", upstream: "http://127.0.0.1:8000/healthz", host: "localhost:8083"
2025/10/28 02:50:36 [error] 20#20: *4 no live upstreams while connecting to upstream, client: 172.17.0.1, server: , request: "GET /f
astapi/healthz HTTP/1.1", upstream: "http://localhost/healthz", host: "localhost:8083"
2025/10/28 02:50:41 [error] 20#20: *5 no live upstreams while connecting to upstream, client: 172.17.0.1, server: , request: "GET /f
astapi/healthz HTTP/1.1", upstream: "http://localhost/healthz", host: "localhost:8083"
2026/02/03 01:48:05 [warn] 11#11: duplicate extension "js", content type: "application/javascript", previous content type: "applicat
ion/javascript" in /etc/nginx/nginx.conf:31
2026/02/03 02:08:05 [error] 13#13: *1 connect() failed (111: Connection refused) while connecting to upstream, client: 172.17.0.1, s
erver: , request: "POST /fastapi/tools/run_ipython_cell HTTP/1.1", upstream: "http://[::1]:8000/tools/run_ipython_cell", host: "127.
0.0.1:49345"
root@3bbacb890183:/var/log#

并且curl测试工具可以用的，如下：
[root@localhost ~]# curl -X POST http://127.0.0.1:49345/fastapi/tools/run_ipython_cell -H "Content-Type: application/json" -H "aut
horization:Bearer 852a0c99bfbb916565ecf7506db4a0d2" -d '{"code": "print(123)"}'
{"meta":null,"content":[{"type":"text","text":"123\n","annotations":null,"description":"stdout"}],"isError":false}[root@localhost ~]

[rayrayraykk]

Could you please try changing the returned domain from "localhost" to "127.0.0.1" here?

agentscope-runtime/src/agentscope_runtime/common/container_clients/docker_client.py

Line 131 in 7045c42

return _id, list(port_mapping.values()), "localhost"

In some environments, localhost may resolve to IPv6 (::1) first, which can cause connection issues if the runtime server is only listening on IPv4. Using 127.0.0.1 forces IPv4 and should make the behavior consistent across hosts.

[aGastronome]

我将ip改为 127.0.0.1 之后还是报相同错误，日志如下：

root@b50bcad39dbd:/var/log# cat agentscope_runtime.err.log
INFO: Started server process [16]
INFO: Waiting for application startup.
Secure MCP Filesystem Server running on stdio
Allowed directories: [ '/workspace' ]
/agentscope_runtime/venv/lib/python3.11/site-packages/pydub/utils.py:170: RuntimeWarning: Couldn't find ffmpeg or avconv - defaultin
g to ffmpeg, but may not work
warn("Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work", RuntimeWarning)
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
INFO: Started server process [12]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
root@b50bcad39dbd:/var/log#
root@b50bcad39dbd:/var/log# cat nginx.
nginx.err.log nginx.out.log
root@b50bcad39dbd:/var/log# cat nginx.err.log
2025/10/28 02:50:27 [warn] 19#19: duplicate extension "js", content type: "application/javascript", previous content type: "applicat
ion/javascript" in /etc/nginx/nginx.conf:30
2025/10/28 02:50:31 [error] 20#20: *1 connect() failed (111: Connection refused) while connecting to upstream, client: 172.17.0.1, s
erver: , request: "GET /fastapi/healthz HTTP/1.1", upstream: "http://127.0.0.1:8000/healthz", host: "localhost:8083"
2025/10/28 02:50:31 [error] 20#20: *1 connect() failed (111: Connection refused) while connecting to upstream, client: 172.17.0.1, s
erver: , request: "GET /fastapi/healthz HTTP/1.1", upstream: "http://127.0.0.1:8000/healthz", host: "localhost:8083"
2025/10/28 02:50:36 [error] 20#20: *4 no live upstreams while connecting to upstream, client: 172.17.0.1, server: , request: "GET /f
astapi/healthz HTTP/1.1", upstream: "http://localhost/healthz", host: "localhost:8083"
2025/10/28 02:50:41 [error] 20#20: *5 no live upstreams while connecting to upstream, client: 172.17.0.1, server: , request: "GET /f
astapi/healthz HTTP/1.1", upstream: "http://localhost/healthz", host: "localhost:8083"
2026/02/03 06:28:12 [warn] 11#11: duplicate extension "js", content type: "application/javascript", previous content type: "applicat
ion/javascript" in /etc/nginx/nginx.conf:31
root@b50bcad39dbd:/var/log#

root@b50bcad39dbd:/var/log# ps aux | grep uvicorn
root 12 0.2 0.0 103020 86648 ? S 06:28 0:01 /agentscope_runtime/venv/bin/python3 /agentscope_runtime/venv/bin
/uvicorn app:app --app-dir=/agentscope_runtime --host=0.0.0.0 --port 8000
root 27 0.0 0.0 3332 1512 pts/0 S+ 06:38 0:00 grep uvicorn
root@b50bcad39dbd:/var/log#

[rayrayraykk]

Which image are you using? Is that the latest tag?

[aGastronome]

这是我修改localhost 为 127.0.0.1后，重启服务，并重新生成的容器id。

[rayrayraykk]

I mean the docker image. docker ps

[aGastronome]

[root@localhost conf]# docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
alias-amd64 v1 7ec574565a55 2 weeks ago 1.35G
redis 6.2 6d62d3a4c126 5 weeks ago 106MB
agentscope-registry.ap-southeast-1.cr.aliyuncs.com/agentscope/runtime-sandbox-alias latest 5c7d62e0c14a 3 months ago 3.19G
agentscope/runtime-sandbox-alias latest 5c7d62e0c14a 3 months ago 3.19G
nginx 1.26.3 64e5ac93d424 12 months ago 192MB

[aGastronome]

[root@localhost conf]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS
NAMES
62161ecf8a10 agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49364->80/tcp, :::49364->80/tcp
runtime_sandbox_container_6zd1h0ip05iri7nrt5dz2uuny
0d81f7267358 agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49363->80/tcp, :::49363->80/tcp
runtime_sandbox_container_ee8eokbpsy85qwhrqlks7bwp5
014d8fbc71d7 agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49362->80/tcp, :::49362->80/tcp
runtime_sandbox_container_7jout7640m0e57icpbrmdnq0o
3c26cd9f752f agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49361->80/tcp, :::49361->80/tcp
runtime_sandbox_container_dbiuwkiuvq4fb6el7y8sogxr8
4b96ad67d574 agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49360->80/tcp, :::49360->80/tcp
runtime_sandbox_container_azgw8kozbaacrxe23mwr4bvno
b50bcad39dbd agentscope/runtime-sandbox-alias:latest "docker-entrypoint.s…" 22 minutes ago Up 22 minutes 0.0.0.0:49354->80/tcp, :::49354->80/tcp
runtime_sandbox_container_66f9e244htxh501lajyuu1dwy
61b2b9182140 alias-amd64:v1 "./start.sh sandbox" 28 hours ago Up 22 minutes
alias-sandbox
6bee1acf6d95 alias-amd64:v1 "./start.sh alias" 4 days ago Up 22 minutes
alias
819107b523e1 redis:6.2 "docker-entrypoint.s…" 11 days ago Up 11 days 0.0.0.0:6379->6379/tcp, :::6379->6379/tcp
alias-redis
20950449d522 nginx:1.26.3 "/docker-entrypoint.…" 2 weeks ago Up 6 days 0.0.0.0:18740->80/tcp, :::18740->80/tcp
nginx

[yangjm-41]

我补充一些信息，调用链条如下：

sequenceDiagram
    participant User
    participant backend as backend:8000
    participant Sandbox as sandbox-server:8001
    participant Docker as Docker Daemon
    participant Container as runtime_sandbox_container

    User->>Alias: 发送 "进入Dify模式"
    Alias->>Alias: MetaPlanner 路由
    Alias->>Sandbox: POST /create_from_pool

    Sandbox->>Docker: DockerClient.create()
    Docker->>Container: 创建容器
    Container-->>Sandbox: 容器ID
    Sandbox-->>Alias: 200 OK


    Alias->>Sandbox: POST /call_tool (run_ipython_cell)

    Note over Sandbox,Container: 问题1：端口映射错误
    Sandbox->>Sandbox: 使用宿主机端口 49164
    Sandbox->>Container: 尝试连接 (超时)

    Note over Sandbox,Container: 问题2：健康检查失败
    Sandbox->>Sandbox: check_health() 期望 200
    Sandbox->>Container: GET /healthz
    Container-->>Sandbox: 403 (认证失败)

    Sandbox->>Sandbox: 等待 60 秒
    Sandbox-->>Alias: ❌ Timeout
    Alias-->>User: ❌ Read timed out

问题 1 是根因，排查可能是此处造成的问题，如下图：