sunchao opened a new pull request, #3691: URL: https://github.com/apache/celeborn/pull/3691
## Why are the changes needed? The HA master client can lose useful leader redirects while bootstrapping or fail over poorly when a suggested leader is stale, unavailable, or forms a redirect chain. In those cases, the client can stop short of trying the remaining configured masters even though a usable master is still available. ## What changes were proposed in this PR? This port brings the HA redirect handling fix from `openai/celeborn#70` onto upstream `main`: - preserve `MasterNotLeaderException` signals through nested causes - follow suggested-leader redirects during bootstrap and failover - fall back to configured masters when redirect setup fails or cycles - advance the configured-master index correctly when endpoint setup throws - add focused regression coverage for bootstrap redirects, chained redirects, redirect cycles, and fallback behavior ## How was this PR tested? - `build/mvn test -pl common -am -Dtest=MasterClientSuiteJ -DwildcardSuites=org.apache.celeborn.common.client.__NoSuchSuite__` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
