Track proxy issues we face with ClickHouse cloud autoscaling #1732
Description
tldr;
During autoscaling (vertical), we expect no impacts on both INSERT and SELECT queries. But we observe sometimes, clickhouse-go throws EOF errors during pod scaling down.
The theories so far
- Proxy "lies" about that underlying connection is good, even upstream CH server is closed
- Client fails to do proper "bad connection" check