KubeCon China 2026, Shanghai: A Recap
Two things got me here, and I kept mixing them up.
Program committee seat = complimentary registration. That’s all it covers. Flight and hotel are on you. So I applied for Linux Foundation Travel Funding separately, on top of the PC seat. Same trip, two different forms, two different problems.
The Travel Funding is basically there for people who otherwise couldn’t make it because of the cost. The Linux Foundation says they get around 1,000 requests a year. For CNCF events, the application asks about your involvement in technology/open source communities and why you want to attend, along with some CNCF-specific questions.
What it covers is pretty specific: coach airfare, accommodation for the event plus one night before, and ground transport to and from the airport. Not food, visa costs, taxis or other non-airport transportation.
They also try not to fund the same person for multiple events in one year, and typically don’t fund the same person two years in a row. So the competition isn’t really the Shanghai pool. It’s everyone else’s year.
There was one detail I kept coming back to: the scholarship has been the Dan Kohn Diversity Scholarship since 2021. CNCF renamed it that year in his memory.
My first KubeCon was in 2019, in Shanghai. It wasn’t called that then.
That’s a small thing, but it made the gap between the two trips feel more concrete. On that first trip, I wouldn’t have seen Dan Kohn’s name attached to the scholarship. Now I’m here applying for it, seven years later, in the same city.
Dan had also been involved in the early KubeCon China story. So the name doesn’t feel like some distant piece of CNCF history anymore. It’s attached to why I’m here this time.
My last Shanghai was 2019, pre-COVID. None of this existed for me then. Not an eight-year gap, exactly, but the city and the reason for being in it both changed underneath me.
And then the sessions started making the connection back to the work I was already doing.
Get Irvi Aini’s stories in your inbox
Join Medium for free to get updates from this writer.
One idea kept repeating: agents are workloads now, not scripts bolted onto the platform. Same obligations as anything else — declarative config, a reconciliation loop instead of cron, tool access set by policy rather than an over-permissioned service account.
Horizon Robotics put it most plainly. They folded their per-workflow agents into one Argo-backed service and got back 1,000+ CPU nodes and 2.4M CNY a month.
A few other talks landed almost directly on top of things already sitting open on my laptop. Cluster API’s declarative underlays. DRA + Kueue and the idea of pooling fragmented GPU clusters. That’s basically the node-rebalancing problem I’m still working through. Envoy Gateway’s dataplane session mapped onto the CLB/TLS work. The impersonation KEP onto the RAM/SCP policy design.
Cloud native hasn’t left the room. It’s just the floor now. For more talk, you can refer to this schedule :D
There was more international traffic at the HAMi booth than expected. AAIF is competing with CNCF not for commits, but for where a new developer’s first interface lands. I was watching the same shift, just at session scale.
Day 1
Breakfast — maintainer/Kubestronaut/Ambassador table. Yujun Liang, Yunjie Zhu (Albert), Yongkang He. Room sorted by hours put in, not by review load.
After lunch — Scholarship Recipient Interviews with CNCF folks where I met fellow recipient Leon Nunes. Less interview, more some QnA with CNCF folks and comparing notes before the interview. He’s answering the same question from the other side.
Last day
Dinner — Miley Fu, Albert (hosting), Abiodun Akanbi, Jeffery Sica, Abigail Casabuena. Room sorted by review load. Everyone at the table spent months reading the same stack of CFPs.
