Ash Berlin-Taylor
Ash Berlin-Taylor
> This PR will add the ideal signal sequence for task runner termination mentioned here: https://github.com/apache/airflow/issues/35474#issuecomment-1801988073. @potiuk Why do we want to send a SIGHUP first? That seems like a...
My default approach here would be to follow what containers to as these days it's an expected pattern, so TERM -> KILL after timeout. and no other signals. I don't...
@kaxil has made a start on a decent TERM->KILL in https://github.com/apache/airflow/pull/44465, though yes, we will need to document this and handle process groups etc.
Yeah don't block on my PR, that is a slow background task for me for 3.1
@folke Any chance of a quick merge on this one? I have tested this locally and it fixes the issue for me on neovim 0.11
Not fixed by #51699 . It's likely something to do with deamon mode closing file descriptors if I were to guess.
`_reopen_std_io_handles` is re-opening them to the sockets already opened in the parent. It is not trying to connect to the "real" tty connected std handles at that point.
> Just curious, why is this prioritized as a low? Is it because people are expected to use: > > `nohup airflow dag-processor > nohup.out &` > > as a...
Sounds like it, yeah @lp-jump
This sounds like "not a bug", if slightly confusing, so I'm going to close this issue. Perhaps a doc update PR might be in order.