dgraph-live-loader is slow when starting two dgraphzero nodes and one dgraph node

[pawanrawal]

Start two dgraphzero nodes and a dgraph instance.

./dgraphzero --id 1
./dgraphzero -id 2 -peer 127.0.0.1:8888 -w w2 -port 8890
./dgraph -memory_mb 10000 -peer 127.0.0.1:8888

Try to load 21 million data with schema

➜  dgraph-live-loader git:(master) ✗ ./dgraph-live-loader -r ../../../benchmarks/data/21million.rdf.gz -s ../../../benchmarks/data/21million.schema

Processing ../../../benchmarks/data/21million.schema

Processing ../../../benchmarks/data/21million.rdf.gz
[Request:      9] Total RDFs done:     9000 RDFs per second:     161 Time Elapsed: 56s 

The starting rate is very low and it keeps on going down. This doesn't happen when only one dgraphzero node is running.

[manishrjain]

Is this because of Raft?

@pawanrawal: Assigned this to @janardhan1993. Reassign it to yourself, if you're already working on this.

[janardhan1993]

Leader changes in zero were not getting updated in state. AssignUids from client was going to wrong server(it was not the leader). zero was throwing error since it was not the leader and client got stuck.