Due to kinda-random route update ordering, the channel stats
are unreliable and may change between test runs.

Jana Babovakova authored 1 month ago and Ondřej Zajíček committed 1 month ago

Also test per-protocol hostanme and router id setting.

Also change one password to be less confusing with the other.

Weirdly, kernel table modification by BIRD is impossibly slower
when querying its contents in parallel by iptools.

There is one full BGP feeder, three simulated uplinks, three ASBRs,
three spine routers and one route reflector.

Everything with IPv6 internal routing only over OSPFv3,
BGP with extended next hops.

There should be really exactly no attribute other than the BGP ones there.

The restart was triggered too soon, causing v3 to sometimes
convert the restart to disable.

Hmm, why we print internal enum type value in 'show route all'?

Add cf-evpn-bgp test case for testing BGP-based EVPN with VXLAN tunnels.

Add 'bridge' and 'vxlan' commands to Netlab to define bridge and vxlan devices.

This reverts commit 8d2cc3f0.
This reverts commit c1eace56.

Update cf-mpls-bgp and cf-mpls-bgp-l3vpn to proper test cases.

MPLS testing is challenging, as labels are allocated dynamically, and
just stripping them away removes almost all useful information and
cannot be used for MPLS tables itself, where labels are keys.

We use three passes:

1) Regular BIRD table tests with stripped MPLS labels, wait for convergence

2) Assign stable labels and prepare maps from MPLS labels to stable ones

3) BIRD and KRT table tests with MPLS labels replaced by stable ones

Note that we require that each node uses a different label range,
therefore labels are globally unique.