Rebalance would affect mutate?

my system is simple designed like this(user push node info to kafka,and another part will consume the message and muate it into dgraph)
I found it often occur error
it makes some data lost.

I have some question:
1.why the leader often change,the total data is 2million. I don’t think the dataset is too large
2.rebalance would affect mutation? when I muate a predicate to one node,at the same time, the relalance would move the same predicate,how does dgraph perform

When dgraph zero move the predicate from one to another, the mutation is not available

it means rebalance and mutate could not work concurrently?

Yes, the dgraph team could provide more detail about this