Skip to main content
Your tools connect over TLS to a single e6data endpoint. You name the cluster you want to query - you never address individual nodes - and e6data routes the request. Cluster upgrades and waking a suspended cluster happen automatically, with no change to your connection.

​
The three protocols

All three are behind one endpoint and run against the same cluster.

​
Routing and concepts

  • Cluster name is the routing key - supplied as a header on gRPC/HTTP, or encoded in the database name (<CLUSTER>/<CATALOG>) on the Postgres path.
  • Catalog and schema select what you query.
  • TLS is always on. The first query to a suspended cluster auto-resumes it (~30–90 s) - HTTP requests wait for the resume; Postgres clients receive a 57P03 retry signal.

​
Operational notes

  • HTTP/2 is required for the gRPC and HTTP paths - use an HTTP/2-capable client and don’t force curl’s --http2 flag.
  • Upgrades are transparent - your connection string never changes.
  • On the HTTP path, reuse sessions via X-Session-Id while polling and paging, and call clearOrCancel when finished.

​
See also

Last modified on September 29, 2026