You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Stephan Ewen (Jira)" <ji...@apache.org> on 2020/09/15 21:44:00 UTC

[jira] [Created] (FLINK-19251) Avoid confusing queue handling in "SplitReader.handleSplitsChanges()"

Stephan Ewen created FLINK-19251:
------------------------------------

             Summary: Avoid confusing queue handling in "SplitReader.handleSplitsChanges()"
                 Key: FLINK-19251
                 URL: https://issues.apache.org/jira/browse/FLINK-19251
             Project: Flink
          Issue Type: Improvement
          Components: Connectors / Common
            Reporter: Stephan Ewen
            Assignee: Stephan Ewen
             Fix For: 1.12.0


Currently, the method {{SplitReader.handleSplitsChanges()}} is passed a queue of split changes to handle. The method may decide to handle only a subset of them and is passes later all remaining changes.

In practice, this ends up being confusing and problematic:
  - It is important to remove the elements from the queue, not accidentally iterate, or the splits will get handles multiple times
  - If the queue is not left empty, the task to handle the changes is immediately re-enqueued. No other operation can happen before all split changes from the queue are handled.

A simpler and more efficient contract would be to simply pass a list of split changes directly and once, for the fetcher to handle. For all implementations so far, this was sufficient and easier.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)