Both HA partners should reply to Lease Query

Problem

A lease query sent to both partners of an HA load balanced pair results in a response from one server and a drop by the other server, based on the load balancing algorithm used to direct non lease query traffic. A response from both servers would reduce the chance of SPOF (single point of failure) events such as network partitions between the lease query client and the active/preferred DHCP server for that client.

Desired solution

Both servers respond when a lease query arrives for an active lease allocated from either partner in the high availability pair. The backup server for a given lease responds some configurable number of milliseconds after the primary server under certain conditions, e.g., the high availability load balanced pair is operating normally, and the server is not the most recent one to have communicated with the client, and the “secs” field in the packet is zero indicating that it is not a retransmission. -> The delay should be a value somewhat greater than the HA lease update latency.