BGP learned route reset on NIC update

[cloudguy]

Hi.  I’m running BGP via FRR in Opnsense 18.7.10.  Have BGP going via FRR, everything works great.  However, noticed that everyone settings are applied to a NIC (eg IP changes, new NIC etc) dynamically learned routes disappear from the opnsense route table.  Have to restart BGP session to get them back.  Is this by design or am I missing something?

[mimugmail]

I'd guess the daemon stops as the Interface is gone. This would also be the case with every other OS.

[cloudguy]

FRR keeps running, BGP session active.  Interface doesn’t go away.  When Interface setting are applied the BGP  learned routes are cleared from the routing table (only static ones persist).  I’m thinking there needs to be a process to refresh routing tables or restart BGP as part of interface “apply” workflow. 

And no, it doesn’t happen on every other OS. 

[mimugmail]

When you bring down an Interface in Linux all Kernel routes go down. Agree on apply action but it's not easy to catch as frr is only a plugin

[cloudguy]

Thanks for this, but unlike an OS, Opnsense is the distribution (not RH / Cent / Freebsd).   I'm thinking there should be a hook built into the FRR package to detect and reload routes on I/F reset / apply.   I've made a note of this in our processes, however other routing / security platforms / distributions have that process link between them.   

Would this constitute a formal feature request or there's another mechanism to post such a request to the OpnSense developers?

[mimugmail]

As this error occurs when reconfiguring an interface it would be a bug report and/or feature request in core repo