Uploaded image for project: 'Flink'
  1. Flink
  2. FLINK-11287

RocksDBListState should use actual registered state serializer instead of serializer provided by descriptor

Attach filesAttach ScreenshotVotersWatch issueWatchersCreate sub-taskLinkCloneUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments
    XMLWordPrintableJSON

Details

    Description

      Currently, when creating a RocksDBListState, the element serializer wrapped by the RocksDBListState is retrieved from the state descriptor:

      static <E, K, N, SV, S extends State, IS extends S> IS create(
          StateDescriptor<S, SV> stateDesc,
          Tuple2<ColumnFamilyHandle, RegisteredKeyValueStateBackendMetaInfo<N, SV>> registerResult,
          RocksDBKeyedStateBackend<K> backend) {
      
          return (IS) new RocksDBListState<>(
              registerResult.f0,
              registerResult.f1.getNamespaceSerializer(),
              (TypeSerializer<List<E>>) registerResult.f1.getStateSerializer(),
              (List<E>) stateDesc.getDefaultValue(),
              ((ListStateDescriptor<E>) stateDesc).getElementSerializer(), // incorrect
              backend);
      }
      

      This is incorrect, since new serializers retrieved from state descriptors have not been checked for compatibility with restored state.

      Instead, the element serializer should be retrieved from the register result, which contains the actual state serializer registered in the state backend for state access (and has already been checked for compatibility / reconfigured appropriately).

      Attachments

        Activity

          This comment will be Viewable by All Users Viewable by All Users
          Cancel

          People

            tzulitai Tzu-Li (Gordon) Tai
            tzulitai Tzu-Li (Gordon) Tai
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Time Tracking

                Estimated:
                Original Estimate - Not Specified
                Not Specified
                Remaining:
                Remaining Estimate - 0h
                0h
                Logged:
                Time Spent - 20m
                20m

                Slack

                  Issue deployment