Currently, there is a single MGS service, and all NodeMap requests are processed by
the MGS, other MDS or OSS nodes will get the copy from MGS and processed it locally
to update the Nodemap status.

In RAFT mode, there will be MGS at multiple nodes, the Nodemap requests will bea
sent to each MGS node to process, then there will a conflict between the processing
of Nodemap request by MGC and the processing by the replicated MGS service.

Another improvement could be store these Nodemap information in the RAFT instead of
the MGS configuration file