Details
Description
When scanning through all n1ql nodes to perform a distributed command (eg get all keys in a system keyspace, distribute a prepared statement to all nodes), clustered monitoring has to load a node map from clustering.
Snag is, clustering doesn't have a full blown map, and queries ns_server every time.
Clustering should have a full map, or at least provide a revision counter method, so that monitoring can host a full map and refresh it when needed.
Attachments
Issue Links
- backports to
-
MB-35670 [BP to 6.0.3] - query clustering to have a revision counter, or query monitoring to cache a full node map
- Closed