1. Provide a summary of the issue:

DNS upstream servers became unhealthy. Client noticed CoreDNS from cluster nodes was flooding the upstream and had to block these requests to allow the upstream server to recover.

1. Proposed title of this feature request

Update CoreDNS and its forward plugin that comes with OCP to support failfast_all_unhealthy_upstreams option

1. What is the nature and description of the request?

With current CoreDNS shipped with OCP, when all upstream are unhealthy it will randomly pick an upstream and still send requests upstream. Client would like to stop all requests going upstream when all are unhealthy. This failfast_all_unhealthy_upstreams option was added in newer versions to add this ability.

1. Why does the customer need this? (List the business requirements here)

Although it does not look like CoreDNS was initial cause for upstream to become unhealthy, it did not help. The DNS issues affected other products and had to stop traffic from OCP clusters to stabilize. Client would like to implement this failfast_all_unhealthy_upstreams option to proactively stop requests going upstream when all servers are unhealthy.

1. List any affected packages or components.

CoreDNS