XiaoHongbo-Hope opened a new pull request, #10424: URL: https://github.com/apache/paimon/pull/10424
## Why A transient DNS lookup failure on the native REST table-token GET currently makes PyPaimon abandon native planning and retry the whole scan through the Python planner. This can turn a short lookup error into a much slower batch, while changing the backend unexpectedly. ## Change Retry only native HTTP GET failures that identify both the `/token` URL and DNS resolution. Use three bounded, jittered delays before retaining the existing fallback. Other planner errors and normal reads are unchanged. ## Validation - Five focused unit tests pass, including bounded retry and non-DNS errors. - In a full 8-rank × 4-worker read-only test, a test-only token DNS exception injected once per process caused the unpatched path to fall back; this patch retried 33 times, completed with 4,611 native plans and 4,611 native reads, zero fallback, and 383.9 samples/s. - Three no-fault complete runs were 389.4 / 383.7 / 390.3 samples/s; no normal-path regression was apparent. The injection validates the Python fallback/retry mechanism, not a real resolver outage. Prolonged outages and DNS failures outside table-token planning remain out of scope. -- 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]
