[jira] [Created] (YARN-12010) [Umbrella] Capacity Scheduler: resolve queue configuration before building queues, with structured validation

["Benjamin Teke (Jira)" <ji...@apache.org>]

Benjamin Teke created YARN-12010:
------------------------------------

             Summary: [Umbrella] Capacity Scheduler: resolve queue configuration before building queues, with structured validation
                 Key: YARN-12010
                 URL: https://issues.apache.org/jira/browse/YARN-12010
             Project: Hadoop YARN
          Issue Type: Improvement
          Components: capacity scheduler
    Affects Versions: 3.5.0
            Reporter: Benjamin Teke
            Assignee: Benjamin Teke
         Attachments: cs-config-resolver.md

Capacity Scheduler queues read and validate their configuration while they are constructed, and inherited values are taken from the live parent queue. As a result, the only way to validate a configuration is to build queues: \{{POST /scheduler-conf/validate}} starts a full throwaway CapacityScheduler per request (about 0.37 s at 1,000 queues in local measurements, against about 0.02 s for the validation engine of this umbrella), errors come back as a single string, and the Capacity Scheduler UI duplicates validation logic in TypeScript.

This umbrella resolves every queue's configuration (explicit key, template, parent, global default, built-in default, derived values) into immutable data before queues are built. Queues are constructed from that data, and a new \{{validate/v2}} endpoint runs the same checks over it without creating a scheduler, returning structured issues that the UI can attach to queues and fields.

Compatibility: no configuration key, default value, store format or public API changes; configurations accepted today stay accepted; legacy REST endpoints are unchanged.
 
Design doc attached.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org