You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Steve Loughran (JIRA)" <ji...@apache.org> on 2019/06/11 11:56:00 UTC

[jira] [Commented] (HADOOP-16358) Add an ARM CI for Hadoop

    [ https://issues.apache.org/jira/browse/HADOOP-16358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16860966#comment-16860966 ] 

Steve Loughran commented on HADOOP-16358:
-----------------------------------------

Yetus is the main test runner we use for the runs; it choreographs the jenkins runs for PRs. So any changes would have to go in there -talk to [~aw]

I don't think anyone has any fundamental issue with ARM, it's just that there's always been ongoing difficulties trying to get the native code to work properly across different CPUs, especially big-endian: you can search for Power, Sun sparc and past ARM JIRAs to see that history.

A big issue here is getting people to care when a change breaks the Arm build: if we don't care that things break then its going to atrophy in exactly the same way that the windows test runs did. And frankly, we aren't being proactive about dealing with jenkins run failures today https://builds.apache.org/job/hadoop-qbt-trunk-java8-linux-x86/

So: not sure what to do here. If someone really wants to do an ARM release, then they're going to have to get involved enough not just to set up the test run, but get it actually working, then own the problem of making sure it still works. Is there someone we can recruit for that

Personal note: not a fan of Zuul. Not going to block anything, but not going to put any effort in either.

> Add an ARM CI for Hadoop
> ------------------------
>
>                 Key: HADOOP-16358
>                 URL: https://issues.apache.org/jira/browse/HADOOP-16358
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: build
>            Reporter: Zhenyu Zheng
>            Priority: Major
>
> Now the CI of Hadoop is handled by jenkins. While the tests are running under x86 ARCH, the ARM arch has not being considered. This leads an problem that we don't have a way to test every pull request that if it'll break the Hadoop deployment on ARM or not.
> We should add a CI system that support ARM arch. Using it, Hadoop can officially support arm release in the future. Here I'd like to introduce OpenLab to the community. [OpenLab|https://openlabtesting.org/] is a open source CI system that can test any open source software on either x86 or arm ARCH, it's mainly used by github projects. Now some [projects|https://github.com/theopenlab/openlab-zuul-jobs/blob/master/zuul.d/jobs.yaml] has integrated it already. Such as containerd (a graduated CNCF project, the arm build will be triggerd in every PR, [https://github.com/containerd/containerd/pulls]), terraform and so on.
> OpenLab uses the open source CI software [Zuul |https://github.com/openstack-infra/zuul] for CI system. Zuul is used by OpenStack community as well. integrating with OpneLab is quite easy using its github app. All config info is open source as well.
> If apache Hadoop community has interested with it, we can help for the integration.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org