The query server works as a pool of processes. Up to "os_process_limit" [1] OS processes are spawned. There is a soft limit config option as well [2]. If the number of idle processes exceeds the soft limit, they will be killed until they are lower than the limit. So the os_process limit effectively defines a configurable parallelism level. On larger cluster sizes those can be increased to something like 1000-2000 for instance from the default of 100.
The query server protocol is described as well [3]. Some newer commands may be missing but otherwise it's fairly current. Cheers, -Nick [1] https://docs.couchdb.org/en/stable/config/query-servers.html#query_server_config/os_process_limit [2] https://docs.couchdb.org/en/stable/config/query-servers.html#query_server_config/os_process_soft_limit [3] https://docs.couchdb.org/en/stable/query-server/protocol.html On Thu, Aug 15, 2024 at 4:12 PM Jamie Bliss <ja...@ivyleav.es> wrote: > Does the current query server handler do any parallelism? eg, pipelining > requests, spawning multiple servers, etc?
