There is a known issue when using queries on CrateDB Cloud data using the Crate AdminUI, or other HTTP-based client, when such queries are expected to exceed 60s.

When query runtimes exceed the cloud provider-specific load balancer timeout, the load balancer's failover strategy will retry the query on the next upstream CrateDB node. This will happen without the client noticing it. The consequence is that such long-running queries may consume all available CPU and hang indefinitely, since large numbers of such queries will run in parallel. We therefore do not recommend using the Crate AdminUI for CrateDB Cloud clusters if your query is expected to exceed 60s (which is the minimum default timeout period for our cloud providers).

In such cases, we recommend avoiding usage of the Crate AdminUI. Instead, we recommend use of a PostgreSQL client for such long-running (i.e., non-idempotent) queries, e.g., JDBC for Java, psycopg2 for Python, pgx for Go, etc. This is an issue originating from the side of the cloud provider and we are working on a solution.

Did this answer your question?