Multiple frontend management in cta
Problem to solve
As we are moving toward GRPc, going away from xrootd 5 HA model, we need to think about how we will manage multiple frontends for:
- different protocols: SSI, GRPC, kerberos for admins?
- high availability
If admins cannot pass commands to a frontend how do they make sure that the frontend is healthy as an eoscta instance endpoint?
Something like cta-admin dr ls for the frontends where we could check ctaVersion, configuration, status, heartbeat... would be nice and consistent.
We need also to be able to regularly ping the frontends using cta-admin and a grpc token too as IPv6 issues in EOS showed us that unidirectional heartbeats are not good enough to ensure that communication works the other way around... (cta-admin version using a GRPC token?).
HA model needs to be thought about: all frontends in 1 DNS name and if one is not OK the GRPC client movs to the next one?
Stakeholders
operators, users for HA model