[
https://issues.apache.org/jira/browse/FLINK-12773?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16874749#comment-16874749
]
Alex commented on FLINK-12773:
------------------------------
More recent failures:
[https://api.travis-ci.org/v3/job/551260640/log.txt]
[https://api.travis-ci.org/v3/job/551265054/log.txt]
[https://api.travis-ci.org/v3/job/551266042/log.txt]
[https://api.travis-ci.org/v3/job/551267985/log.txt]
Also, when I try to download the file manually in a browser (via a direct url
or via Kafka downloads page), the browser downloads up to 100% progress, but
then waits for a while to finish download. I'm using Chrome browser.
[~Zentol], in your PR for the new framework, you are using Java specific
hashing as file key. What do you thing about using more neutral algorithm, for
example md5 hash sum. This way I can try to implement downloading as a script
that would be compatible with your download mechanism.
> Unstable kafka e2e test
> -----------------------
>
> Key: FLINK-12773
> URL: https://issues.apache.org/jira/browse/FLINK-12773
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Kafka, Tests
> Affects Versions: 1.9.0
> Reporter: Dawid Wysakowicz
> Priority: Major
>
> 'Kafka 0.10 end-to-end test' fails on Travis occasionally because of
> corrupted downloaded kafka archive.
> https://api.travis-ci.org/v3/job/542507472/log.txt
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
