OK, there's a try-with-resource to close the TableLoader in
FlinkInputFormat [1]. so we don't have to do the extra try-with-resource
in PR 2051 ( I will close that).
Under my host, I did not reproduce your connection leak issues when
running TestFlinkInputFormatReaderDeletes. Did you have an
I use the try-with-resource pattern in the FLIP-27 dev branch. I saw this
problem in Flink tests with the master branch too (although less likely).
With the FLIP-27 dev branch and an additional DeleteReadTests, it almost
happened 100%.
Also, the Spark module (in the master branch) also has this fl
It could be that there are two separate flaky test issues with not
releasing connections in Flink and Spark. I don't think that the
HiveCatalog code has been changed much recently, which would point toward
problems elsewhere.
I think one good reason to use HiveCatalog is to catch problems like the
Hi Piyush,
You might want to consider having a separate partition stats file for each
partition spec. That way, each stats file contains just one partition
struct type and you can keep the struct unmodified. There is a way to
convert a partition struct to a string (PartitionSpec.partitionToPath) b