CTRL+C does not work in error cases

[amotl]

Hi again,

thanks once more for conceiving this excellent package, we are about to explore its most recent features. Along the lines, we wanted to report that an application using the package does not react to KeyboardInterrupt / SIGINT. Maybe you can do something about it?

With kind regards,
Andreas.

Appendix

Problem

$ python example.py
[2025-10-24T23:21:31Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
[2025-10-24T23:21:32Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
[2025-10-24T23:21:33Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
[2025-10-24T23:21:34Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
[2025-10-24T23:21:38Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
^C^C^C^C^C^C^C^C^C^C^C^C^C^C[2025-10-24T23:21:44Z ERROR r2d2] error connecting to server: Connection refused (os error 61)
^C^C^C^C^C^C^C^C^C^C^C^C^C[2025-10-24T23:21:57Z ERROR r2d2] error connecting to server: Connection refused (os error 61)

Observations

At some point, the application will finally fail and trip.

pyo3_runtime.PanicException: Could not create a pool of connections: Error(Some("error connecting to server: Connection refused (os error 61)"))

30 seconds default timeout?

$ time python example.py
[...]
real	0m30.367s
user	0m0.182s
sys	0m0.058s

[surister]

30 seconds default timeout?
Yes the default timeout is 30s

{
        Builder {
            max_size: 10,
            min_idle: None,
            test_on_check_out: true,
            idle_timeout: Some(Duration::from_secs(10 * 60)),
            max_lifetime: Some(Duration::from_secs(30 * 60)),
            connection_timeout: Duration::from_secs(30),
            error_handler: Box::new(LoggingErrorHandler),
            event_handler: Box::new(NopEventHandler),
            connection_customizer: Box::new(NopConnectionCustomizer),
            thread_pool: None,
            reaper_rate: Duration::from_secs(30),
            _p: PhantomData,
        }
    }

It could be configured if needed.

About the CTRL+C it's unfortunately not trivial, r2d2 errors are not programming errors until the timeout is reached since it's trying to reconnect to the server, meanwhile while rust is running python doesn't get the chance to check for signals, there is the API py.check_signals but I'm not sure where should I call that withing the stack

[surister]

These are the solutions I could think of

1. Make max_timeout configurable
2. Create a handler for ctrl+c signal and kill the process, but users wouldn't be able to graciously catch the ctrl+c

[surister]

Alternatively we could apply this patch sfackler/r2d2#152, rebuild the pool ourselves and create a custom error handler that if flagged, stops the execution.

[... 5 minutes later]
But what if the program is running ok and I still want to stop it