So, i try to: signal.signal call directly after/before app, with and without py-call-osafterfork and py-master-check-signals options - not work
signal.signal call in postfork function, with and
without py-call-osafterfork option - not work
I try to use alarms on log, but that not emits, neither when i try to
use example from documentation. I try to:
cmd: echo ...
cmd: curl ...
log:
- not work
Probably, that is the problem with gevent which i use.
I found some not-documented uwsgi.atexit, which work well!
It looks, that I solve the same problem as
https://github.com/unbit/uwsgi/issues/680 describe. So it could be
documented, because we need the PID of exited worker. For another
solution, i could need some other resources from worker. That is, why i
can't use hook-as-user-atexit.
The better could be atexit decorator, which is what i try to found
first.
Ondra
V Wed, 22 Aug 2018 14:34:49 +0200
Ondrej Tuma <konfere...@webjet.cz> napsáno:
> Hi there,
>
> i try to create custom metrics for prometheus with official
> python prometheus_client library. I use uwsgi with gevent and Flask,
> with more workers.
>
> So i need to handle, when worker shut down to call
>
> prometheus_client.mark_process_dead(worker_pid)
>
> But i can't found another solution that log_alarm which will check
> message like:
>
> Brutally killing worker 7 (pid: 22)...
>
> Please, have you anybody some better solution, or way, how can i
> check, that worker was kill. I try to use signal.signal handler but
> that not work.
>
> Thanks a lot.
>
--
Ondřej Tůma <mc...@zeropage.cz>
https://ipv6.mcbig.cz https://keybase.io/mcbig twitter: mcbig_cz
pgpO3eIqnpXSE.pgp
Description: Digitální podpis OpenPGP
_______________________________________________ uWSGI mailing list uWSGI@lists.unbit.it http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
