KubeVirt at KubeCon NA
Andrew Burden
KubeCon NA 23 was held at McCormick Place West in Chicago over 4 days, Nov 6-9, with the first day covering all project meetings and co-located events. The CNCF has not yet posted their ‘KubeCon in numbers’ but I’ve seen attendance of 9000 (down from 10K in Amsterdam).
The KubeVirt party in attendance consisted of project maintainers Ryan Hallisey, Vladik Romanovsky, and Andrew Burden. Unfortunately Fabian Deutsch, who was scheduled to present at a number of sessions, was unable to attend at the last minute; fortunately Vladik was able to fill in for him and did an admirable job.
Project Meeting
First up Monday morning, we had our KubeVirt Project Meeting, not at the venue but nearby in a Hilton meeting room. Attendance was low, and aside from three maintainers from KubeVirt, we had representatives from SK Telecom and Kasten by Veeam.
The smaller group meant it was a more casual conversation. Our agenda covered the recent releases, SIGS update, and rootful VM deprecation. We also covered what a useful roadmap would look like. The folks from SK Telecom raised some questions about upgrading their clusters and live migration, migrating ephemeral and persistent storage, and enquired about a CSI best practice matrix (not as a recommendation, but to help guide users a bit to find something to suit their needs).
OpenShift Commons
One of the KubeCon colocated events on Monday. Of note was the CRI-O graduation celebration video, and our own Fabian Deutsch was recognised as a Community Hero.
V1.1 Release
We had decided to delay the v1.1 release by one week so as to coincide with KubeCon, which opened us up for promotional opportunities through the CNCF. This included coordinated blogs with the CNCF and additional social media promotion through their accounts, and an interview with TheNewStack.
It was also a useful talking point at the kiosk as we made it our kiosk splash page and were able to build on a lot of the v1.0 conversations (since we had a v1.0 maintainers talk and a v1.0 project update) to talk about the straight-out-of-the-oven fresh new features. More on this below but the promotion of the release helped drive new folks to our kiosk later in the week to hear more about the project.
TheNewStack Interview
Vladik and Ryan were interviewed by TheNewStack to discuss the v1.1 release. There's no recording yet but everyone was happy with how it went.
Maintainer Talk
Vladik and Ryan presented "We Finally Released v1.0. Now what?" and covered some KubeVirt history as a lead-in to some of the big changes made to align with Kubernetes: our release cadence and the push towards growing SIGs within KubeVirt. They then elaborated on some of the features for v1.0 (with a sprinkle of 1.1) and our aims and expectations for the project in the immediate future.
There were about 15 mins of questions asking about the philosophy of running big VMs in Kubernetes (big/legacy VMs are part of our original use case), Red Hat’s relationship with KubeVirt and how that works with OpenShift Virtualization (Red Hat contributes to KubeVirt; KubeVirt encourages/needs contributions from everyone in our community), can you stack KubeVirt with Kubernetes on top of something like OpenStack as a coherent ecosystem to satisfy broad use cases (KubeVirt is part of the cloud native ecosystem), do we run databases on KubeVirt (we have end users running databases on VMs in KubeVirt), is it as production ready as libvirt/KVM? (KubeVirt already has some very mature virtualization features but because it works in a Kubernetes context there are some gaps or differences to how that feature might work, but KubeVirt is production ready)
The estimated attendance was 80-100.
Project Kiosk
Similar to KubeCon EU 23 we ran a full-time kiosk in the CNCF project pavilion across all three days. Andrew tries to track individual organisations that we talk to at the booth and he counted ~80, slightly less than the ~100 we saw in Amsterdam. This almost tracks with the difference in total attendance for the conference. One big difference was that in Amsterdam most of those conversations happened on the first day, whereas in Chicago it was more evenly split across all three, even the last day which is quite short comparatively speaking. A number of people mentioned that they had heard about KubeVirt in the keynotes and wanted to learn more about it, which may have helped drive the third-day traffic.
The general feeling is very positive, possibly even more so than when we met in Amsterdam. One noticeable difference was that the requirement for bare metal nodes was less of a deal breaker. Whether this is a trend tipping away from cloud or just a different population is hard to say.
Alongside the “We want to learn more about this KubeVirt” and the “We use KubeVirt and we love it” folks, most questions or sounds of interest came around migrating existing VMs from VMWare and OpenStack, and also the multitenancy possibilities of the Cluster API Provisioner KubeVirt project. People also love to hear about the power of CDI and what it can do for them, and are relieved about the design and application of our storage and network agnosticism.
A couple of takeaways from conversations with folks, mostly to cater for new users:
We should post our four use cases for KubeVirt on our website to make them more visible.
New users are wary of what they might need to do for networking, and how adding multus might affect their existing network configuration. Maybe we could add a simplified intro to networking for VMs?
Users interested in importing VMs want to know more about how this affects their storage and networking. Maybe we could add a short 'Importing VMs to KubeVirt FAQ'?
And Finally
A big hello and thanks for dropping by to any and all who dropped by the kiosk to ask questions, thank us on behalf of the project, or who just wanted to say g'day and learn more about the project.