Closed brantburnett closed 6 years ago
Based on feedback from @ingenthr this pull may be based on some incorrect or misunderstood information I was given. It seems that using a reverse proxy has some potential problems during failure situations, and DNS SRV is the recommended course. Though I still need to research options for DNS SRV forwarding in Kubernetes, I have found using the ExternalName service type to be useful there.
Motivation
Some clusters are configured to bootstrap via a single reverse proxy (or load balancer) using HTTP. This configuration has the advantage of making it easy to configure disparate microservices to connect to the cluster, and also allows full cluster swaps using XDCR during cluster maintenance. The SDK Doctor currently throws errors and warnings connecting to a cluster using this configuration.
Modifications
Add a new flag
--reverse-proxy
which:Results
Clusters configured using the more advanced reverse proxy bootstrap will now test cleanly if the new flag is supplied. The default behavior is unchanged.