- Good day,
I have new set of clusters to be deployed in environments where I do have needs for SDN (vxlan-evpn) and loopback routes for CEPH while preparing for IPv6 only networks. Though I did got it going at one stage (IPv4 only), there are still missing pieces/processes that I'm stumbling on that either I misread/misunderstand the documentation, or simply aren't documented nor implemented yet
So the problems I currently experience that isn't documented, or that might be perhaps in 8.2 but haven't found yet myself so that is why I have to state I'm testing/FAFOing on PVE 8.1.10 at present
0. there are missing documentation especially w.r.t. what is happening in what part of the configuration processes, so yes, SDN is still HBD/caveat emptor for anything not simple easy
- When having IPv6 endpoints, the BGP router-id (which is a IPv4 quad-octet) gets filled in with the IPv6 IP, and then FRR fails on that - difficult to see/find as I've not found a CLI why to run/debug other that
systemctl restart frr ; systemctl status frrand/orvtysh -c "wr t"checking the configs of/etc/frr/frr.conf - This brings me to
/etc/frr/frr.conf.local- gleamed from forum posts, not documented, and still "buggy"/lacking w.r.t. what is/isn't loaded- For one, I have some stuff I want added to
router isis routernetnamebut havingrouter isis routernetnamein frr.conf.local, gives this error below, and having it without the routernetname, there aren't an error, but the generated frr.conf has thatrouter isisstanza separately, and thus frr throws errors/doesn't properly load
- several IPv6 keywords/configs not yet "accepted" (yes, could add the code perhaps myself, but this post is to document for others to be aware of ), like:
Reason I'm "needing" this is that I wanted thespf prefix-priority critical ACL_LISTfor reasons attempts to faster converge ISIS to not have othe errors during SDN Apply
- For one, I have some stuff I want added to
- I'm missing hooks (perhaps they are there, couldn't yet gleam from the source code where/how) debug and/or add delays during the SDN apply process.
- Specifically where/what is doing a
/sbin/ip get route 2cff:ffff:feed:feed:1:0:5:3- that fails during reloading, as that is a loopback on one of the nodes, that ISIS haven't yet converged with - yes, it is the EVPN/VTEP and BGP peer end point, but I've been unable to find the "culprit", as I need to add some delays before that, or see why FRR was restarted, instead of reloaded.
- Specifically where/what is doing a
- There is a "Warning" that should've been an error that does a FRR restart, instead of reload, and doing that restart (using something like systemctl) it misses the errors generated by FRR for the config
- ISIS net id isn't properly parsed in GUI nor during config parsing and frr.conf generation. See above warning that should be error - only picked this up as I was rechecking frr.conf & vtysh's wr t output.
- THERE ARE NO Factory Reset for SDN, so difficult to go back to a "clean" state
- there is also
/etc/pve/sdn/.version,/etc/pve/sdn/.running_configand/etc/network/interfaces.d/sdn, as well as the removal (manual/reboot) of the interfaces/bridges to have a "clean" state - recreation of
/etc/frr/frr.confis somehow dependant on something else, but because I don't have a debug, nor had time to re-re-re-read the code, insights into the flow/process of the SDN Apply, I can't say what is or isn't done where that might be failing to also fix.
- there is also
-
ifreload -anot IPv6 only ready w.r.t. vxlan?
