When there is no master node yet elected (or got into this state because of minimum master nodes), or the cluster has not recovered yet, a proper failure should be raised on all actions. Rest status code should be 503.

Original Request:

When a node in a cluster reboots, there is a brief period of time while that node is searching for a master that that node can 404, even if discovery.zen.minimum_master_nodes equals 2:

% sudo service elasticsearch restart
...
time passes
...
% curl http://localhost:9200/index/type/id
{"error":"IndexMissingException[[index] missing]","status":404}
% curl http://localhost:9200/_cluster/health
{"error" : "MasterNotDiscoveredException[]", "status" : 500}

Since this is the same error message you get when the there's actually no index, I need to check with /_cluster/health to see if the cluster is in this startup phase to tell if this is a temporary error or not. It'd be preferable if instead elasticsearch could just raise a separate error so I don't need to do that.