Lower Level Capabilities in Network Faults

[prb-aerospike]

Describe the solution you'd like

1. Fault-injection on the network ports:
   At present, mangle only supports injecting the network faults such as delay, packet drop on entire network interface. It would be nice to specify the particular port(s) of an interface as well to inject these faults rather than on entire network interface.

2. Three-way Partitioning:
   Mangle supports network partitioning between 2 units - endpoint(s) and target hosts that can split the distributed systems cluster into 2 partitions. 3 partitions or partitioning between 3 entities will be useful to test the resiliency of distributed system especially with the cluster running in the AWS environment where cluster nodes could be present in the different availability zones.

Describe alternatives you've considered
For now, we are using other chaos engineering tools to inject the common network faults on the particular port(s) of a network interface.