c-thiel commented on issue #14762: URL: https://github.com/apache/iceberg/issues/14762#issuecomment-3631212073
The strange thing is that Spark seems to use the IO for the table also for the Checkpoint. In Lakekeeper we see a very clear log that states: ``` Received a table specific sign request for table xxx with a location s3://xxx/warehouse/019ae372-7eed-75c3-bca2-b35bcd38def0 that does not match the request URI https://s3.us-west-2.amazonaws.com/xxx/dlh2/spark-checkpoints/iceberg-streaming/bronze_to_silver/my_table/sources/0/offsets/0 ``` According to @koombal the same happens with STS - Spark tries to use the vended credentials returned from a LoadTableResponse to use for the Checkpoint IO as well - which is again forbidden because the Checkpoint is not in the tables location. @nastra do you know how a Table specific FileIO could leak out and be used to store a checkpoint? I tried browsing the Java code but nothing caught my eye. -- 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]
