ajantha-bhat commented on PR #16945: URL: https://github.com/apache/iceberg/pull/16945#issuecomment-4790750209
@Baunsgaard: My thinking with the reusable planner was to avoid copying the same JVM policy into every workflow and risking drift between Java/Spark/Flink/Hive/Kafka/Delta. In terms of raw runner minutes, the planner jobs should be tiny compared to the JDK 21 jobs this removes. This PR adds a few short plan jobs, but removes many heavy test/build jobs from the default PR path, so the net runner-minute savings should still be large. That said, I agree the plan job is not free: it is still a scheduled runner job and it gates fan-out in each workflow. Since the original problem is ASF shared-runner pressure and queueing, I can see why avoiding even small extra jobs is attractive. So I think the trade-off is: - reusable planner: less duplication and one JVM policy to maintain - inline expression: no extra jobs and no additional fan-out gate I am gonna update this PR to go with inline expressions instead of a single JVM policy to make it more faster. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
