`pino.transport.pipeline` may have lost logs about pino HOT 10 OPEN

Can you reproduce without pino-syslog? I think the bug is there and not in pino proper.

More specifically, pino-syslog does not implement the close() callback (https://github.com/pinojs/pino-abstract-transport?tab=readme-ov-file#buildfn-opts--stream) to tear down the destination stream properly. As a result, whatever is "in flight" there will be lost.

from pino.

@mcollina

I tried adding a close callback function, but it did not prevent or "delay" the thread(Worker)'s exit.

In other words, the thread had already closed before I could do anything

https://github.com/pinojs/thread-stream/blob/main/lib/worker.js#L147-L153

from pino.

You linked the logic that catches unhandledRejections.. how is that related?

from pino.

You linked the logic that catches unhandledRejections.. how is that related?

Sorry, the link above is incorrect, and in the above text, I mistakenly wrote the main process as a thread

from pino.

Please add reproductions.

from pino.

thank you for your prompt reply

My confusion stems from the fact that once Node.js enters the exit event, even if we add a ref to the worker stream, the main process will not wait for any active resources and will terminate immediately after the autoEnd function completes. Is there a mechanism within Pino itself that can prompt the main process to wait?

from pino.

The transport functionality I’m aiming to implement is as follows:

1. To collect logs passed down from upstream and, depending on their level, write them to different files (similar to how pm2 does it).
2. Then, to synchronize these logs with a syslog server, which is why I need pino-syslog.

like : #318

from pino.

My confusion stems from the fact that once Node.js enters the exit event, even if we add a ref to the worker stream, the main process will not wait for any active resources and will terminate immediately after the autoEnd function completes. Is there a mechanism within Pino itself that can prompt the main process to wait?

Yes, pino has its logic there to handle that case, while pino-syslog doesn't use it.

from pino.

Yes, pino has its logic there to handle that case, while pino-syslog doesn't use it.

I’m seeking guidance. How can I ‘delay’ the exit process?

from pino.

My confusion stems from the fact that once Node.js enters the exit event, even if we add a ref to the worker stream, the main process will not wait for any active resources and will terminate immediately after the autoEnd function completes. Is there a mechanism within Pino itself that can prompt the main process to wait?

Yes, pino has its logic there to handle that case, while pino-syslog doesn't use it.

Add autoEnd : false ?

from pino.