Short description:

Indexer had an incorrect logic to identify active indexer nodes during multi-service rebalance (at least kv and index service part of rebalance). This could cause a potential downtime on new clients created leading to failures in index creation, builds, or other DDL operations.

Problem Description:

When a rebalance is performed, new nodes which are to be added to the cluster show up in nodes endpoint with status as `active`. If an active node has `index` service, we consider it for port translation in our iterations. However, the service ports may not show up in nodeServices endpoint until the service rebalance is started which will happen after KV rebalance. Hence any calls to translate port during the KV rebalance could lead to port translation errors for nodes which come after the new nodes in the nodes endpoint. This means until KV rebalance, index creations, builds, alters could have a potential downtime on new clients.

Problem resolution:

Update the information used by TranslatePort to use the nodeServices endpoint