Hi, We observed MetadataFetchFailedException during executorLost if spark.speculation enabled.
looks like something out of sync between original task on failed executor and its speculative counterpart task? is it a known issue? please let me know if you need more details. Thanks, Renyi.