🎉

@cpuguy83 @olljanat @vdemeester

Nice. Now I "just" need study how to install and configure Jenkins...

A couple of things to bear in mind 🐻

1. To ensure this doesn't get merged and bit-rot we should have a clear plan to enable a Jenkins Multi-branch pipeline for moby/moby. This will involve stopping leeroy sending webhooks to the old Jenkins freestyle jobs

2. To do a rebuild of a PR for a certain job, one will have to use the Jenkins UI. The Multibranch pipeline plugin will create a "branch" each PR. You just need to hit "build", deselect the builds that you don't want and then you are done. In theory this could be automated so the slack/irc style rebuild commands still work.

3. I've used the checked in powershell script for Windows RS1 CI vs. the one that is in-lined in the Jenkins job as it makes the Jenkinsfile over 1000 lines 😨 This is the same script used for RS5. @jhowardmsft do you think that this would be ok?

4. There are some inconsistencies between how the bundles.tar.gz is created and how container cleanup happens. It would be nice to fix this but it could be done in a follow up.

5. I haven't seen a green build just yet. @thaJeztah I might need your help figuring out what is just flaky vs. what's a genuine failure.

@dave-tucker first of all. Thanks for sharing this config with us. Just to have it visible simplifies many things as we don't need to guess.

1. To ensure this doesn't get merged and bit-rot we should have a clear plan to enable a Jenkins Multi-branch pipeline for moby/moby. This will involve stopping leeroy sending webhooks to the old Jenkins freestyle jobs
2. To do a rebuild of a PR for a certain job, one will have to use the Jenkins UI. The Multibranch pipeline plugin will create a "branch" each PR. You just need to hit "build", deselect the builds that you don't want and then you are done. In theory this could be automated so the slack/irc style rebuild commands still work.

I don't know how others want to use this config but I would prefer to move logic to Jenkinsfile to generic bash scripts. First example of that kind of approach is on olljanat@936fc81

On that way:

• contributors are able to test/debug most common CI related things locally without need to setup Jenkins.
• we can allow contributors to enable CircleCI to forked repos by including super simple config like this to Moby olljanat@a1dcada (you can see log from test run on here )
• IMO, we don't necessarily need migrate Jenkins to use Jenkinsfile but instead of just export it to here on documentation purposes as it should need updates very rarely.

3. I've used the checked in powershell script for Windows RS1 CI vs. the one that is in-lined in the Jenkins job as it makes the Jenkinsfile over 1000 lines fearful This is the same script used for RS5. @jhowardmsft do you think that this would be ok?

I have understood that that logic have been in Jenkins on past and it was moved to PowerShell to simplify local debugging (seen comment on some old PR).

4. There are some inconsistencies between how the bundles.tar.gz is created and how container cleanup happens. It would be nice to fix this but it could be done in a follow up.

Fixed already on example commit above 😉

5. I haven't seen a green build just yet. @thaJeztah I might need your help figuring out what is just flaky vs. what's a genuine failure.

At least most common flaky issues are listed on #37306

Codecov Report

Merging #38565 into master will increase coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #38565      +/-   ##
==========================================
+ Coverage   36.43%   36.44%   +<.01%     
==========================================
  Files         613      613              
  Lines       45847    46481     +634     
==========================================
+ Hits        16706    16941     +235     
- Misses      26849    27155     +306     
- Partials     2292     2385      +93

Updated to add a 3 hr timeout to avoid builds staying in a pending state forever.

@olljanat I like the bash scripts ❤️

3. I've used the checked in powershell script for Windows RS1 CI vs. the one that is in-lined in the Jenkins job as it makes the Jenkinsfile over 1000 lines 😨 This is the same script used for RS5. @jhowardmsft do you think that this would be ok?

@dave-tucker - yes, hack\ci\windows.ps1 is the right script.

I pushed some changes to add a customWorkspace for the Windows builds as I missed this when extracting info from the current Jenkins builds.

Now I think this is good to go 🎉

See: https://jenkins.dockerproject.org/job/dave-tucker/job/jenkinsfile-test/

I've seen it pass at least once on each set of tests.

java.lang.NoSuchMethodError: No such DSL method 'addFailedStep'

Good catch @cpuguy83 - I've just updated this PR to fix that!

Exception from GitHubCommitStatusSetter for windowsRS5-process: org.jenkinsci.plugins.scriptsecurity.sandbox.RejectedAccessException: Scripts not permitted to use method hudson.plugins.git.GitSCM getRepositories

Oh, somehow I grabbed logs from an old run...

Oh well, powerpc has the same error on the last run. Seems to be triggered when a particular pipeline step fails.

Let's merge this; the file won't be used for now, as follow-up changes are needed to change the CI config and so on 😇

@cpuguy83 @thaJeztah thanks!

Exception from GitHubCommitStatusSetter for windowsRS5-process: org.jenkinsci.plugins.scriptsecurity.sandbox.RejectedAccessException: Scripts not permitted to use method hudson.plugins.git.GitSCM getRepositories

This is a quick fix - I just need to add that method to the whitelist in Jenkins...
But if I did that now, then it would start setting commit status for real, which we don't want until we actually do the migration to Jenkins Multibranch Pipelines 😉