Open iljarotar opened 1 month ago
Actually the metal-bmc address to connect is queried from metal-api by describing the partition where the machine resides and the takes Partition.MgmtServiceAddress
https://github.com/metal-stack/metal-api/blob/master/cmd/metal-api/internal/service/v1/partition.go#L8
This means we must expose the metal-bmc on a Address which terminates on both management servers and is routed with ECMP to both sides
Actually we only connect to one metal-bmc instance to forward console access. If one management server is down console access is not possible anymore. Possible solution: loadbalance between slice of metal-bmcs.