You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Till Rohrmann (JIRA)" <ji...@apache.org> on 2019/05/13 14:22:00 UTC

[jira] [Created] (FLINK-12502) Harden JobMasterTest#testRequestNextInputSplitWithDataSourceFailover

Till Rohrmann created FLINK-12502:
-------------------------------------

             Summary: Harden JobMasterTest#testRequestNextInputSplitWithDataSourceFailover
                 Key: FLINK-12502
                 URL: https://issues.apache.org/jira/browse/FLINK-12502
             Project: Flink
          Issue Type: Bug
          Components: Runtime / Coordination, Tests
    Affects Versions: 1.9.0
            Reporter: Till Rohrmann


The {{JobMasterTest#testRequestNextInputSplitWithDataSourceFailover}} relies on how many files you have in your working directory. This assumption is quite brittle. Instead we should explicitly instantiate an {{InputSplitAssigner}} with a defined number of input splits. 

Moreover, we should make the assertions more explicit: Input split comparisons should not rely solely on the length of the input split data.

Maybe it is also not necessary to capture the full {{TaskDeploymentDescriptor}} because we could already know the producer's and consumer's {{JobVertexID}} when we create the {{JobGraph}}.



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