Reads of medium to large size data sets fails on nodes with replica shards

when running dsbenchmark on a node with replica shard, the READ operation on a medium-sized list fails.

I try to many reads (items in a list) on a remote leader (I.e. in my 3-node test cluster I issued the read on 10.194.126.98 or 10.194.126..99, which are replica nodes), the operation never finishes. The dsbenchmark READ test dumps a 10k-element list into the data store, and then tries to read the elements one by one. A list of 1,000 items works fine. A list of 10,000 items does not work. Even with the 10k-item list, I can see the list items through RESTCONF in the leader node (issue a REST read request on 10.194.126.97), but trying to do the programmatic read from a remote node does not work.

To reproduce, install dsbenchmark and run the attached script first on the leader node and then on replica nodes with the following command line:

./dsbenchmark.py --host 10.194.126.98 --txtype SIMPLE-TX --inner 1 --optype READ --warmup 1 --runs 3 --total 10000 --ops 10 100