taptap tcp connection issue?

[litinoveweedle]

Hello,

as you probably know I do provide Python mqtt wrapper and package your taptap as Home Assistant addon.

I am using Python subprocess.popen to read from the taptap stdout and stderr. Time to time I observe strange behavior where no new messages are produced by taptap after some time. The process is still running but not logging anything. If I restart taptap it will immediately starts login new messages. TCP connection is listed as in state ESTABLISHED, but there are no data flowing between taptap and the Modbus converter (validated by tcpdump).

The issue can be as well caused by sketchy Modbus convertor implementation. (i.e. improper Modbus TCP session handling etc.), but it would be nice if taptap can detect such state.

I also checked your code, especially the function observe in the main.rs and I am not sure, if your code handle correctly all edge cases - for example TCP session timeout/reset/connection loss to the converter. To my use case it will be sufficient to implement taptap to exit with error code, so no reconnect logic is necessary. I tested this theory by restarting Modbus converter, but taptap process not detecting that - connection is still listed as ESTABLISHED and no data are flowing.

Thank you.

[willglynn]

I saw you opened then closed #15 adding TCP keepalive support. Are you still investigating whether that addresses the issue?

[litinoveweedle]

Hello, thank you for reply.

I did open it by mistake, as it was untested and it has compilation issue. I corrected it later and now I have working patch with the TCP keep alive support, but I think it will not help completely. The reason is, that problem is probably caused by the poor Modbus converter implementation.

When testing the patch I saw, that there are normally many messages sent by gateway even when the optimizers are down (night). So I would propose to implement some logic to quit the taptap process when there are no messages received within the given configurable timeframe. It doesn't make sense to implement TCP keep alive, if there is always sufficient traffic to keep session open.

What do you think?

[litinoveweedle]

@willglynn so what do you suggest?

I can push you PR for TCP keepalives, but this will probably not fix issue nicely. This is as because there is always enough traffic to not allow TCP session to timeout. TCP keepalive might detect when Modbus convertor loose session, but it would be better to implement functionality it in the receive loop in the taptap.

The question is - I have before this try done 0 lines in Rust, so my work is painfully slow and probably with terrible quality. Are you planning to maintain this project? I have not seen any commit past the initial one. I can right now open at least 2 other PR - one with Github action to build and release binaries and second with fixing CLI paramter bug for port. But I have not seen any activity from your side so far in the codebase... So what are your plans?

[litinoveweedle]

@willglynn may I please ask you for a reply?

[willglynn]

@litinoveweedle Sorry, I've been traveling (and still am). I am happy to maintain this project – to preserve its functionality, to review code submissions, to publish updates – but do not currently have the time or imperative to sit down and develop new features. By all means, open PRs with the expectation that they could be merged and released.