Hi Charles,

I did, yes. My strategy is to break the large test set down into several
waves. Actually, I have 5 total waves for unit tests: small, medium,
large-1, large-2, large-3. They run in parallel and so the total time for
feedback is much faster than the Jenkins model... at the cost of pretty
high resource usage on branch-2*. It's possible that we'll run into issues
with Github's generosity with the free tier runners, but we'll have to
cross that bridge with INFRA if/when it comes up.

-n

On Fri, Feb 6, 2026 at 4:39 PM Charles Connell via dev <[email protected]>
wrote:

> Thanks for your work Nick,
>
> When exploring Github actions myself, I found that they might struggle
> to run the unit tests in the hbase-server module. Any single action
> step is limited to 6 hours runtime, and running these tests can take
> longer than that on the action runner, in my experience. Did you work
> around this somehow?
>
> Charles
>
> On Fri, Feb 6, 2026 at 10:29 AM Nick Dimiduk <[email protected]> wrote:
> >
> > Heya team,
> >
> > We now have static analysis, build, and unit test coverage via GHA
> implemented for PRs against all branches. In the absence of comments,
> you'll find the detailed Yetus table reports in the "Action Summary"
> section of the checks interface. Implementation details are attached to
> HBASE-29838. As before, shout if you find something broken.
> >
> > As of this effort, all of our PR work is being done in GHA. All of the
> Jenkins infrastructure is still present and intact, but I have disabled the
> job in the Jenkins UI [0]. If you're having problems with the new tooling
> and want to restore the Jenkins job, just flip that switch back to enabled.
> >
> > There's more jobs to be converted, I've not filed any additional Jiras
> yet. Please feel encouraged to tackle something if you're feeling so
> inclined.
> >
> > Thanks,
> > Nick
> >
> > [0]: https://ci-hbase.apache.org/job/HBase-PreCommit-GitHub-PR
> >
> > On 2026/01/15 14:35:50 Nick Dimiduk wrote:
> > > Hi team,
> > >
> > > I've added GitHub Actions to run Yetus general checks directly on pull
> requests. This has been backported to all active branches (master,
> branch-3, branch-2, branch-2.6, branch-2.5).
> > >
> > > What's changing:
> > >  - PRs will now trigger a "Yetus General Check" workflow that runs
> checkstyle, javac, pylint, shellcheck, and other static analysis checks
> > >  - Results appear as GitHub status checks and PR comments (same as
> Jenkins)
> > >  - Old bot comments from previous runs are automatically hidden to
> reduce noise
> > >  - Uses the same scripts, Dockerfile, &c. that we use for Jenkins
> > >
> > > What's NOT changing:
> > >  - Jenkins precommit builds continue to run as before
> > >  - Your PR workflow remains the same
> > >  - Unit/integration tests still run on Jenkins, not GitHub Actions
> > >
> > > JIRAs: HBASE-29787, HBASE-29829
> > >
> > > Give us a shout if anything breaks for you. Also, feel free to pick up
> migrating other jobs, if you want to (I've not filed any subsequent
> tickets). It's a little bit tedious and it seems that you need push rights
> to the origin github repo to introduce a new workflow (I'm not sure about
> editing an existing workflow).
> > >
> > > Thanks,
> > > Nick
> > >
> > > On 2024/10/02 09:22:51 Nick Dimiduk wrote:
> > > > Heya,
> > > >
> > > > I'd like to take the community temperature on migrating our build
> infra
> > > > from the ci-hbase.a.o Jenkins instance to something built on GitHub
> > > > Actions. I have several reasons that justify this proposal.
> > > >
> > > > As some of you may know, our community funding has reduced and we
> will no
> > > > longer be able to sustain the current fleet of build infrastructure.
> So,
> > > > one motivation for this proposal is cost-cutting: I think that we'll
> be
> > > > able to operate at lower costs if we can migrate to a
> provisioned-as-needed
> > > > model of consumption.
> > > >
> > > > My second reason is an optimistic appeal to a larger contributor
> base. I
> > > > suspect that if we can modernize our infrastructure then we will
> increase
> > > > the pool of contributors who might be able to participate in this
> area. I
> > > > believe that GH Actions (and systems like it) is more prevalent in
> the
> > > > industry than Jenkins, which means that more people already have
> experience
> > > > with the platform and more people will feel compelled to offer
> support to
> > > > an OSS project that uses the platform as a means of growing their own
> > > > skillset and as a means of bolstering their CVs.
> > > >
> > > > Dove-tailed into reason two is reason three: I believe that there is
> a
> > > > large community of folks who are developing GitHub Actions on its
> > > > marketplace. We would effectively open ourselves up to more
> off-the-shelf
> > > > offerings and those offerings would be in our hands directly. By
> contrast,
> > > > I don't think there's as much development in Jenkins plugins, and the
> > > > process of adding a new plugin to our Jenkins instance requires
> filing an
> > > > INFRA ticket.
> > > >
> > > > These are my motivations. I'm still not clear on what's possible yet
> for
> > > > ASF projects. I have filed an INFRA ticket, requesting whatever is
> > > > necessary for us to start an experiment. Indeed, I believe that
> there are
> > > > some major limitations on the current implementation provided by the
> ASF,
> > > > and as far as I can tell, only one project with a build footprint
> that
> > > > resembles HBase has pursued this effort. I've catalogued the
> applicable
> > > > information that I've found so far on that issue.
> > > >
> > > > https://issues.apache.org/jira/browse/INFRA-26170
> > > >
> > > > Thanks,
> > > > Nick
> > > >
> > >
>

Reply via email to