[ 
https://issues.apache.org/jira/browse/FLINK-8890?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16389399#comment-16389399
 ] 

ASF GitHub Bot commented on FLINK-8890:
---------------------------------------

Github user zentol commented on a diff in the pull request:

    https://github.com/apache/flink/pull/5653#discussion_r172807512
  
    --- Diff: 
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CompletedCheckpoint.java
 ---
    @@ -290,16 +289,14 @@ public static boolean checkpointsMatch(
                Collection<CompletedCheckpoint> first,
                Collection<CompletedCheckpoint> second) {
     
    -           Set<Tuple2<Long, JobID>> firstInterestingFields =
    -                   new HashSet<>();
    +           List<Tuple2<Long, JobID>> firstInterestingFields = new 
ArrayList<>();
     
                for (CompletedCheckpoint checkpoint : first) {
                        firstInterestingFields.add(
                                new Tuple2<>(checkpoint.getCheckpointID(), 
checkpoint.getJobId()));
                }
     
    -           Set<Tuple2<Long, JobID>> secondInterestingFields =
    -                   new HashSet<>();
    +           List<Tuple2<Long, JobID>> secondInterestingFields = new 
ArrayList<>();
    --- End diff --
    
    initialize with `second.size()`


> Compare checkpoints with order in CompletedCheckpoint.checkpointsMatch()
> ------------------------------------------------------------------------
>
>                 Key: FLINK-8890
>                 URL: https://issues.apache.org/jira/browse/FLINK-8890
>             Project: Flink
>          Issue Type: Improvement
>          Components: State Backends, Checkpointing
>            Reporter: Aljoscha Krettek
>            Assignee: Aljoscha Krettek
>            Priority: Major
>             Fix For: 1.3.3, 1.5.0, 1.4.2
>
>
> This method is used, among other things, to check if a list of restored 
> checkpoints is stable after several restore attempts in the ZooKeeper 
> checkpoint store. The order of checkpoints is somewhat important because we 
> want the latest checkpoint to stay the latest checkpoint.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to