1. A Slap in the Face: Deleting the Repo, the Images, the Docs
On Ray’s last day at QingCloud as the project’s initiator, the frontend, docs, and images of the open source project KubeSphere were deleted, and users received an [Announcement on the Adjustment of the KubeSphere Open Source Project] through various channels.
Even if the company’s management was deeply divided over KubeSphere, and even if it was hostile and targeted at certain people, there was no need for it to be this crude and this fast. All one can say is that this was handled very childishly, like a child throwing a tantrum.
You can stop maintaining an open source project. You can just host it on free GitHub Pages. So many copies have already been forked. What was the point of deleting all this?
2. KubeSphere and Me
I cherish every work experience I have had, because they make up who I am now. I do not want others to slander the places I have worked. Things always have a good side and a bad side, but what matters more is the side that fits. Slander comes only from an incomplete understanding, not from something being truly bad.
Between 2019 and 2021, I took part in the development of KubeSphere at QingCloud. At the time the team was spread across Beijing, Shanghai, and Wuhan, and at its largest numbered more than 30 people.
The sense of vigorous upward momentum at KubeSphere was exhilarating. Every operational metric was growing, the community was getting more active, and quite a few online and offline events were held. There was something new to do every day, and the boundaries of our work kept expanding.
When I left, I even wrote a blog post, The Road to Open Source for a ToB Startup - KubeSphere, to record how I felt at the time.
Later, at a new company, I found out they were evaluating KubeSphere. The single sentence “he worked on KubeSphere” made the recognition that experience brought unmistakable.
3. QingCloud Held KubeSphere Back
On July 8, 2020, SUSE announced it would acquire Rancher Labs for 600 to 700 million dollars. Right then I realized that KubeSphere was very valuable β it was just that QingCloud never brought out its value.
The contradiction lies here: QingCloud hoped KubeSphere would drive IaaS sales, but Kubernetes lowers the value of IaaS. Kubernetes frees development and operations from all kinds of IaaS products, letting you design for Kubernetes without caring which cloud vendor’s services you use.
If KubeSphere could have been spun off, compatible with every major cloud vendor’s IaaS, and fought for the entry point to application-layer development and deployment, that would have been a better choice. On the user side, KubeSphere already had a solid foundation and an active community; giving developers a certain free quota, scheduling IaaS resources across the whole network to build a multi-cloud application platform, capturing developer and small-to-medium business traffic online, and selling an offline edition to large enterprises. On the cost side, with a heterogeneous multi-cloud platform plus Kubernetes’ high-density deployment, KubeSphere’s operating costs would have held up.
But KubeSphere could not escape QingCloud’s shackles. It was constantly busy with pre-sales and after-sales support, busy serving QingCloud’s customers, and it missed the chance to develop independently. By the time the infrastructure trend shifted from cloud native to AI Infra, QingCloud could no longer see KubeSphere’s excess value and kicked it aside.
4. The Future of the Community Edition of KubeSphere
Working full-time on development at KubeSphere came with one very bad experience: the team really treated you only as an employee, not as a community member, with no well-formed system for joining and leaving. When you left, you were deleted straight from the GitHub team, and I have never seen a departing employee file a single PR or issue with the KubeSphere community.
For an outward-facing project like this, departing employees are a very important resource β they are sparks scattered outward. A word of approval from a former employee at another company can bring great positive influence; conversely, the negative influence is just as great.
QingCloud closing the source of KubeSphere is nothing more than rushing to push users toward the enterprise edition and the cloud edition in response to pressure from the capital markets; there is nothing unreasonable about that, and I understand it. At this point, if departing employees gathered together and, under the project’s open source license, maintained a new community edition, that could be a viable plan.
There are two reasons for this. First, I do not want to see the project simply disappear, with years of our investment scattered to the wind. Second, many clusters may already be using KubeSphere-related components, and there is a need to cover the downside and prevent panic.
Just as the open source ecosystem is getting better and better, KubeSphere got kicked and became a martyr. Truly a shame.
