dramaticlly commented on issue #4168: URL: https://github.com/apache/iceberg/issues/4168#issuecomment-1045055044
So essentially what we observed is that the Flink S3 FileIO failed to upload the data files due to file not exist, but the missing data file is incorrectly tracked in iceberg manifest. So the subsequent query against the partition will fail as the claimed data file cannot be found. Our iceberg was setup on AWS S3 with versioned bucket and we can confirm that the - data file never get uploaded, no version exists of given path - the data file is being tracked in iceberg and we do see such non-exist data file in iceberg metadata query This result in some nasty behavior where we need to reconcile the manifest state based on what exists in underlying file system, which is not possible. So we had to drop the partition as temporary work around. But I think it would be helpful to understand why do we run into the issue where iceberg commits before s3 confirm the data file is uploaded? In the meantime, we are trying to see if we can have a simple repro of the issue CC @szehon-ho -- 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]
