exabgp worker does not emit any debugging log meesage

[Yang Youseok]

Currently in exabgp_peer_worker.py, 'decorated_restart' decorator steal exabgp logger initialization. But the problem is there is no way to set exabgp log level if it does not restart with 'restart' flag.

So whatever user add config level, exabgp debugging log does not emit at all and it makes hard to find solution with even tiny misconfiguration like invalid local_address.

[Thomas Morin (tmmorin-orange)]

The log level is supposed to be forced here:
https://github.com/openstack/networking-bagpipe/blob/master/networking_bagpipe/bagpipe_bgp/engine/exabgp_peer_worker.py#L102

This should work independently of the restart step in exabgp unmodified Logger.

I'm not sure what causes the problem you are seeing. I'd need to investigate more.
I suspect that the way we monkey patch exabgp to drive logs to oslo_log stopped working with some change in exabgp.

Could you indicate which version of networking-bagpipe you are using, and with which version of exabgp ?

[Yang Youseok (ileixe)]

4.0.5 for exabagp, stable/queens for networking-bagpipe.

I dig the exabgp code base a little bit, and I found the way in the code you mentions do not change the logger log level in exabgp, but do change the kind of metadata when exabgp's print function use.

So, if syslog logger level is not changed (in restart step that I said), nothing happens with syslog._debug() function.

I think the way of exabgp logger is kind of weird though, maybe we networking-bagpipe should follow
it.

[Thomas Morin (tmmorin-orange)]

> I think the way of exabgp logger is kind of weird though, maybe we networking-bagpipe should follow it.

Feel free to make practical proposals.
My current stand was: in networking-bagpipe, we want everything logged through oslo_log for consistency, so we need to bypass the specific exabgp logging.

I'll also try to dig more why the current code is not actually controlling the exabgp log level.

[Thomas Morin (tmmorin-orange)]

I've tried again in the dev env I have available right now with me, which is not queens, but pre-rocky with exabgp 4.0.6, and the feature works as designed in this context.

Just to make sure: are you aware that '-d' is not sufficient to have all the logging of all components, because some components (and exabgp is part of them) have their logging level controlled differently:

https://github.com/openstack/networking-bagpipe/blob/stable/queens/networking_bagpipe/bagpipe_bgp/bgp_daemon.py#L51-L60

To override these settings and enable debug for exabgp, you'll have to use the --log-config-append option. This behavior is not specific to bagpipe-bgp, but common to many openstack components.

If you do this and still don't have exabgp debug, then this means we'll need to reproduce and dig on a queens environment.