[BUG]: Wrong error when no connection to APM socket is possible

[cmitz]

Module Version(s)

1.5.0

Bug Report

Our nginx container runs with the Datadog APM socket file mounted as volume. It is also explicitly configured to use these. However, the container still logs this when it cannot connect to that socket.

2025/02/20 16:08:26 [error] 94#96: nginx-datadog: [error code 4] Error occurred during HTTP request for Remote Configuration: Error sending request with libcurl (Couldn't connect to server): Failed to connect to localhost port 80 after 0 ms: Couldn't connect to server

The expected behaviour is that it logs it can't communicate over that socket, instead of saying it cannot connect to localhost port 80

Reproduction Code

No response

Error Logs

No response

Operating System

No response

[sureshkothapally]

any update on this?

[zacharycmontoya]

Hi folks, I investigated this and what's happening here is that the Datadog module is relying on the configured http client to handle http operations, and in this particular case it is relaying the error message from the underlying curl implementation. From these lines of code in the core tracing library, you can see that in order to use unix sockets (instead of TCP/IP), we must form the HTTP request with a valid host (localhost) as required by curl:

https://github.com/DataDog/dd-trace-cpp/blob/bb4891d5d4fab50be313b4aba5021cb01fffe76e/src/datadog/curl.cpp#L384-L385

As a result, when curl reports an error about the request, it is reporting the dummy host localhost and port 80. Based on my investigation, I believe fixing this error message would require an upstream change to fix the log message in curl itself.