Hello, We are currently developing a RichParallelSourceFunction<> that reads from different FileSystem dynamically based on the configuration provided when starting the job.
When running the tests, adding the hadoop-s3-presto library in the classpath, we can run the workload without any issues. However, when using our cluster, in Kubernetes, we have this exception <https://github.com/apache/flink/blob/master/flink-core/src/main/java/org/apache/flink/core/fs/FileSystem.java#L514> raised when getting the file system, we have enabled the fs using the ENABLE_BUILT_IN_PLUGINS env variable. The code that initialize and get the FS is the following one, the configuration contains s3 keys to provide runtime credentials instead of using the cluster ones: FileSystem.initialize(configuration, null); FileSystem.get(<s3_path>); We strongly believe that the issue comes from the plugin manager, and that we probably are missing something when initializing/configuring the FS. But after some days of debugging/testing we still can't figure it out, have you any idea on what would Thanks in advance for your help, Gil
