As we have come to expect from their conventions this past week’s KubeCon has been unique. In previous years there has been snow in Austin, rain in San Diego, this year is no different with a virtual conference. Being my first virtual conference experience, even with the lack of people, excitement, and swag, I was able to real-time attend the keynote and many sessions. 

To follow is a summary of announcements and releases that captures my highlights from all sessions I attended.

Keynotes:

Day 1

There is something unique to how KubeCon does keynotes compared to other industry conferences that I’ve attended. KubeCon uses multiple speakers that provide a wide-ranging experience. In short, it’s all about short snippets of a whole ecosystem in highlight form. Some of the highlights shown during the first day are all about growth and recognition. In the growth segment of the keynote, they discuss statistics around this year’s projects. 

Here are some of my favorite statistics they highlighted in the keynotes:

• There are 150+ Projects in the CNCF.
• 1.3 Billion Lines of code have been written across those projects.
• 97,000 individual contributors across the projects.

The second part of Day 1’s keynote focused on recognition. To summarize, everyone makes a difference, no matter what your title or job. Each of us helped to create the current state of the ecosystem we all enjoy, so for that, we personally thank each and every one of you, there are a few things I found interesting from this initial keynote

• Cheryl Hung gave a talk on Why End Users Are More Than Passive Consumers
• Kubernetes itself is moving closer to that boring stage, that stage where it becomes ubiquitous, and that is a good thing as the exciting things are happening above the base Kubernetes platform. Things like controllers and operator automation are new ways to connect with the existing world and modify things.
• End users and vendors are playing in each other’s roles.

Day 2

With many of the breakout sessions and tracks about halfway completed, I wanted to highlight the KubeCon Keynotes Day 2. It started out with Vicki Cheung giving us a lot of project updates in the Kubernetes project specifically. 

The big news everyone already knew about was K8s v1.19 coming out in November containing a litany of enhancements:

• Generic ephemeral Volumes (more than just emptyDir)
• Ipv6 on windows
• Initial support for cgroups v2
• Block device reached v1 (yay for stability)
• PVC Cloning for backups 🙂
• Immutable Secrets & configmaps
• HPA Controls
• Finer control on scaling up/down and give specific intervals for each
• Scheduling Enhancements
• Run Multiple Schedulers without them stepping on each other.
• Tain Based Evictions is GA
• PodTopologySpread moving to beta (Logical AZ spreading)
• Ephemeral containers with alpha `kubectl alpha debug` that can be added onto a pods context later.
• FlowSchema’s allow api-servers from being overloaded with priorities for specific api endpoints.
• Node Topology Manager allows for more efficient use of resources that need consolidation on the same node.
• Kubectl diff is stable now 🙂

Briana Frank gave a talk called, The Kubernetes Effect – Igniting Transformation in Your Team. She helped show that Cloud Native is really about speed, and getting your company to be able to deploy +1000 updates a day to production, without stress. She gave a few good options for greenfield like, allowing the teams to re-evaluate the current path and do a full rewrite if it has the ability to increase velocity by orders of magnitude and improve rather than work inside the existing mature/old codebase. Another good takeaway was that using design thinking and hashing out the plan first with all the stakeholders (and ensuring everyone is using the same plan for the final output), get everyone that has a horse in the race involved. This might be a no brainer for KubeCon attendee’s but starting new projects on Kubernetes can reduce the effort moving forwards, and is a huge win. 

Briana shared was the Idea of Automation Culture and how it takes hold quickly and grows fast things like:

• Chat-ops for the win! Use a slack bot for gate approvals from dev to production.
• Write Documentation if you have received the same question more than a few times, then you can just hand the next asker the link, and reduce toil.
• Once documentation is mature enough, automate the resolution so you don’t need to manually resolve it next time.

Holly Cummins gave a keynote speech on How to Love Kubernetes and Not Wreck the Planet that was a visual and auditory view of how the world can reduce our impact and not lose anything in the process. She talked on how IBM on average used to have 21 clusters per client last year and that they had been focusing on reducing the cluster sprawl. With a big focus on consolidation and reducing the number of clusters and decreasing zombie workloads in each cluster, they were able to reduce the number of nodes in clusters by 25% or 4000 servers. To do this reduction, it took a lot of effort, but the main takeaways were reducing the number of clusters as more control planes doesn’t equal better implementations. Implementing a leased resources system with auto removal that removed workloads that had expired via a label that you could update to extend the lease every few weeks. In short, automation helps.

Hannah Foxwell from VMware Labs, gave a keynote on the Human Approach to Coping with Complexity, and how we can improve things. She preached people over processes, and how to reduce the Developer and Engineering Toil. Hannah recommends to cap toil at < 50%, if your KPI’s start to see ~50% Toil. Pain goes up substantially from there, so halt features development and focus on Toil Reduction to increase future velocity. Then when you’re back to a relative lower Toil percentage resume improvements.

Stay tuned for Part 2 of our KubeCon 2020 Overview sharing highlight from the 3rd and 4th days. It will be posted on our blog on Monday.