Great experience so far, it "just works". My primary CI was Circle CI until now around which I have developed a lot of custom scripting but seeing this I think I will gravitate towards pre-ci from now on. What I especially like is the overall responsiveness of this Jenkins instance, post-commit Jenkins CI is way slower, loading a lot. This just reacts instantly.
On Wed, Aug 27, 2025 at 11:50 PM Mick <m...@apache.org> wrote: > > tl;dr > I am super excited to share with you all the availability of > https://pre-ci.cassandra.apache.org , made possible from a very generous > donation by Amazon. Using AWS donated credits we now have a fast > open-source CI system for pre-commit testing. This is intended primarily > for committers (and trusted contributors) that do not have access to a > company-private CI system. Contact me off-list if that's you. > > … > This huge donation from Amazon and builds on years of work on our CI and > testing scripts, accumulating in CASSANDRA-18145¹. > > The project now has an AWS account (controlled by ASF Infra), and into it > Amazon has provided a significant amount of credits for the purpose of our > pre-commit testing. > > In the C* project pre-commit CI has been a major hurdle and blocker to new > contributors, unaffiliated committers, and generally getting patches > committed. Too much of our momentum has been led by contributors and > committers employed in companies that can pay to run their own in-house CI > systems. It is not in the slightest their fault, nor their employers' > priority, but it has held the project back from its potential as well as > not being in the proper spirit of an open source project and community. > > Amazon has recognised this and stepped forward as a true champion of our > open-source community, providing both the AWS credits, as well as time in > helping setting it up, testing and documenting it. > > The work has also involved many steps and people the past five years! > since that first mention of a "Stand-alone Pipeline Run"². This work has > gone from having a standalone Jenkinsfile, testing images and scripts that > can be used anywhere, both pre-commit and post-commit, to a Helm chart to > deploy ci-cassandra.a.o Jenkins clones in any k8s cluster of your > choosing. This has also included getting our CI pipeline generating > aggregated test result artefact files that can easily be attached to jira > tickets, and archived permanently. These objectives you can see listed in > CASSANDRA-18137³. > > For all this special thanks goes out to Richa Mishra, Nishant Barola, > Aleksandr Volochnev, Brandon Hsieh, Himanshu Jindal, Steve Mayszak, Dmitry > Konstantinov, Maxim Muzafarov, Berenguer Blasi, David Capwell, Brandon > Williams, Ekaterina Dimitrova, Henrik Ingo, Shankar Ramaswamy, Josh > McKenzie, Stefan Miklosovic, Gavin McDonald and no doubt others I've > forgotten! Yeah, it feels like almost everyone, it has been a journey! > > It is not perfect: there is (always) room for improvement; and for now I > think next steps should be around further time+cost optimisations and > pipeline configurability. I'm hoping folk can jump in and help with > these. > > What is provided today is a way for everyone anywhere to immutably > provision a ci-cassandra.a.o clone on k8s. You avoid any need to install, > know or operate Jenkins (you do not even have to ever once open the Jenkins > UI – see below) and you do not need to wait more than a few hours to run > the "post-commit" pipeline with 140-200k tests. The correlation to > pipeline time is now how many nodes your k8s can scale to (ofc limitations > exist, but those fine details are other convo), and with the cluster > scaling down to one node you get a cost-effective pay-what-you-use > deployment. https://pre-ci.cassandra.apache.org runs on EKS, and another > instance for DataStax employees ( http://ci-cassandra.infra.datastax.com > ) has been in use on GKE for over a year now. We have a CI system that is > relatively ops-free, has the same turnaround time as CircleCI and costs > over an order of magnitude less. > This has been designed for downstream C* products to be able to re-use, > and indeed it is a setup that any project can easily use. > > The setup introduces repeatability in how we test. The Jenkinsfile, the > docker images and the test scripts, are the same everywhere, including > locally. Reproducing obscure failures in post-commit should be easier. > > If you would like to set up your own ci-cassandra clone, the docs for both > GKE are found in-tree under .jenkins/k8s/ ,and the docs for EKS coming soon. > > > There is a command line interface that can be used for both provisioning > as well as everyday CI use. > Some examples: > ``` > # Install the ci-cassandra helm chat in your current k8s context, creating > the ci-cassandra clone > > .build/run-ci --only-setup > > # Run your current branch through CI with the skinny profile. > # Waits til completion, and creates ci_summary*.html & > results_details*.tar.xz files to attach to jira > > .build/run-ci --profile skinny > > # If you don't have kubectl access to the cluster, just specify the > jenkins url and user > > .build/run-ci --profile skinny --url pre-ci.cassandra.apache.org --user > myuser > ``` > > > Huge huge praise and thanks to Amazon here: they really deserve it, they > have taken this over the finish line; but the opportunity deserves to > repeat all our gratitudes to all the very significant > ci-cassandra.apache.org donations that help us keep the project running: > NetApp, DataStax, Huawei and iland. > > > ¹) CASSANDRA-18145 – Run entire Cassandra Jenkins in any k8s (immutable > provisioning) > ²) https://lists.apache.org/thread/ch768l8y95npno9o9y130fjob1h675z8 > ³) CASSANDRA-18137 – Repeatable ci-cassandra.a.o (Epic) > > > >
