Unauthorized NCs should be prevented from joining the CBAS cluster

1. Create a cluster with 1 KV node.
2. Add a CBAS node to the cluster.
3. Create a default bucket.
4. create 100000 documents
5. Setup Cbas:

create bucket default_bucket with {"name":"default"};

create shadow dataset default_ds on default_bucket WHERE `profession`="doctor";

connect bucket default_bucket

6. Create index on query also to cross check the items.
CREATE INDEX NoName ON default(profession)
7. Check for items in CBAS, they are not same as in default bucket:

select count(*) from default_ds

• Sometimes the ingestion process is very slow due to DCP connections getting dropped.
• Sometimes the ingestion stuck at some point.

On one set of centos machines it always failed for many previous builds also: 172.23.109.48, 172.23.99.9
Logs:
https://s3.amazonaws.com/bugdb/jira/DCP_connections_dropping/collectinfo-2018-06-15T080157-ns_1%40172.23.109.48.zip
https://s3.amazonaws.com/bugdb/jira/DCP_connections_dropping/collectinfo-2018-06-15T080157-ns_1%40172.23.99.9.zip

Passing on another set of centos machines: 172.23.109.223, 172.23.109.224
Passing on Debian machines: 172.23.96.89, 172.23.96.90

Script to run the test:

/opt/jython/bin/jython -J-cp Couchbase-Java-Client-2.5.6/*:jsch-0.1.54.jar:doc_ops.jar testrunner.py -i ${ini_file} -t cbas.cbas_partial_rollback.PartialRollback_CBAS.test_ingestion_after_kv_rollback_create_ops,cb_bucket_name=default,cbas_bucket_name=default_bucket,cbas_dataset_name=default_ds,items=100000,where_field=profession,where_value=doctor,compression_mode=active