These are chat archives for Exa-Networks/exabgp

26th
Apr 2017
imperalix
@imperalix_twitter
Apr 26 2017 20:40
I want to manage what IP addresses exabgp advertises. I was going down the route of adding the ip to my healthcheck config. However, I see that when i send a SIGUSR1, exabgp doesn't reload the healthcheck config.
Thomas Mangin
@thomas-mangin
Apr 26 2017 20:44
Happy to discuss it tomorrow GMT
imperalix
@imperalix_twitter
Apr 26 2017 20:46
Sounds good, thanks.
Justin
@JustinAzoff
Apr 26 2017 20:52
@imperalix_twitter you should look into using a 'process' section in your config
the process just needs to output to stdout announce/withdraw statements
imperalix
@imperalix_twitter
Apr 26 2017 20:53
@JustinAzoff Maybe that's my problem. I have a default process that has run .../healthcheck.py -F healthcheck.cfg
and i'm updating healthcheck.cfg with my health check command and vip
Justin
@JustinAzoff
Apr 26 2017 20:54
ah.. then it sounds like you need to be HUPing your healthcheck process, not exabgp
imperalix
@imperalix_twitter
Apr 26 2017 20:54
i tried both
Let me try again to make sure I'm not doing something stupid.
Justin
@JustinAzoff
Apr 26 2017 20:56
ah I see.. the healthcheck.py that comes with exabgp? I don't think it ever re-reads its config
imperalix
@imperalix_twitter
Apr 26 2017 20:57
Yeah, the one that comes with exabgp
Justin
@JustinAzoff
Apr 26 2017 20:57
ah!
 USR1   Causes exabgp to reload the configuration.
 USR2   Causes exabgp to reload the configuration and restart any forked processes.
according to the man page
so give SIGUSR2 a try
imperalix
@imperalix_twitter
Apr 26 2017 20:58
let me try that
Thanks, not sure how I missed it.
Justin
@JustinAzoff
Apr 26 2017 20:58
I run my process with a simple shell script wrapper, so I can restart it for code updates or crashes automatically
mostly just a while true;do real-process;sleep 1;done
imperalix
@imperalix_twitter
Apr 26 2017 20:59
How do you health check?
Justin
@JustinAzoff
Apr 26 2017 21:01
oh, I don't use this for health checks, I use it for injecting null routes for bad things
imperalix
@imperalix_twitter
Apr 26 2017 21:02
makes sense
Justin
@JustinAzoff
Apr 26 2017 21:02
it does need some better monitoring.. there's 2 things that have gone wrong.. the process itself gets hosed (or I break it, like the time I accidentally added an extra ; to the commands)
or the bgp peering gets hosed somehow
Ideally I'd like to just monitor the number of advertised routes that the remote side sees.. still trying to get access to that sort of thing
imperalix
@imperalix_twitter
Apr 26 2017 21:04
I hear you! I have a todo to have a blackhole detector that alerts, testing oob. Since i'm trying to not to blackhole stuff.
:)
Justin
@JustinAzoff
Apr 26 2017 21:04
something weird happened with peering/exabgp last week causing things to not sync up
and with crazy amounts of ECMP that was a fun one to figure out why things were sort of kind of not being blocked
heh, I'm currently blackholing 170k routes.. on purpose :)
imperalix
@imperalix_twitter
Apr 26 2017 21:11
Looks like that did the trick. Thanks. Now on to the next problem.