- Fixing issue with nil pointer de-reference due to bad design of mesh
manager.
- Going forward all references to GetSelf should be depracated. It
introduces a race condition when leaving a mesh network
- Ability to specify WireGuard keepalive in the CLI formatter
- Ability to specify publicly routeable endpoint
- Ability to specify whether to advetise routes into the mesh,
and whether to advertise default routes.
- Ability to run a command pre up and post down
- Ability to be a client in one mesh and a peer in the other
- Added dev card to specify different sync rate, keepalive rate per
mesh.
- Routes are not being removed despite being withdrawn from the
configuration.
- Best path routes are not shared across interfaces
- Bug in consistent hashing wrong parameter passed caused by
refactorings.
Keepalive is based on per mesh and not per node.
Using total ordering mechanism similar to paxos to elect a leader
if leader doesn't update it's timestamp within 3 * keepAlive then
give the leader a gravestone and elect the next leader.
Leader is bassed on lexicographically ordered public key.
- Added mesh-to-mesh routing of hop count > 1
- If there is a tie-breaker with respect to the hop-count use consistent
hashing to determine the route to take based on the public key.
Only consider clients who have updated their config when synchronising
with peers. Consider a dead time where we don't have a handshake and
a prune time when we remove them from the WireGuard configuration.
Removed a client's grpc endpoint value. Client's aren't publicly
available so there is no need for a client's gRPC endpoint.
Also changed a node ID's to their public key. A node id's public
address is an issue for mobility of clients as their endpoint
is subject to change
Hashing the interface and using ephmeral ports so that the admin doesn't
choose an interface and port combination. An administrator can alteranatively
decide to provide port but this isn't critical.
Stubbing our WireGuard components so that I can use docker/podman
network_mode=host. This is much more efficient than the docker/podman
userspace network.
