Week Ending March 4, 2018

Community Meeting Summary

Chuck Ha kicked off the Community Meeting with a demo of Sonobuoy, a tool for running Kubernetes’ recently broadened conformance tests. Heptio staff wrote Sonobuoy to make it easier to run the tests, so that contributors will run them more often and prevent unexpected API breakage.

SIG updates this week included Big Data, where Anirudh Ramanathan explained several of the projects, mostly external, to make data mining and analytics work better on Kubernetes. First, Apache Spark 2.3 was released that day, with a top feature of Kubernetes support, and 3 SIG members are now ASF committers. He also mentioned trying to build a Kube-native version of Apache Airflow, work on HDFS support, and kube-arbitrator for big data resource management.

SIG Storage has released a number of alpha features in the last few releases, including Container Storage Interface (CSI), local storage support, mount propagation, and volume protection. They will be working on getting those to GA over the next few releases, also on topology-aware storage for 1.11. Multicluster had less encouraging news; first, Saad Ali said that they would not be fixing existing bugs in Federation V1 as work moves to Federation V2. The latter is in its own WG, which is currently in the “brainstorming” phase and plans to have a federation API which is different from the standard single-cluster API. They are also working on Cluster Registry for multi-cluster metadata, and Kubemci, a CLI for Ingress in multiple clusters.

Announcements: Phil Wittrock said that the project will stop distinguishing between Maintainers and Owners. Meet Our Contributors is Wednesday, contributor participants wanted. And Kubecon.EU registration prices go up March 9.

Release Schedule

We are now in week 9 of 12 for 1.10 and in Code Freeze. During Freeze, only merges for the upcoming release are accepted so that SIGs will focus entirely on stability, bug-fixing, and technical debt. Contributors should have already submitted PRs for docs last Friday, and all users should test out the betas. We are also collecting known issues.

Feature Work

Since we are not in Code Freeze, feature commits all target 1.10.

• The Kube-scheduler can now manage extended resources, which both expands scheduler pluggability and breaks some older default behaviors.
• DaemonSets, StatefulSets, and the garbage collector now all have Conformance Tests.
• CRI can now configure Windows containers using WindowsContainerResources, and exposes Docker image label data.
• client-go now has exec-based support for third-party authentication providers.
• Both GCE and AWS now support block device volumes.
• Nodeport IP Addresses are now configurable, supporting clusters with multiple networks.
• The IPVS feature gateway is now enabled by default.
• Nodes can now request auth tokens on behalf of pods running on them.
• Users can now specify the GID of containers.

Deprecated / Compatibility

• We are now automatically lower-casing Pod names when generating static Pods based on hostname.

Version Updates

• Dashboard to 1.8.3.

Other Merges

• Kubelet now updates Systemd status.
• Failed Pods are no longer calculated towards HPA totals.
• CRD Finalizers can execute actions during deletion.
• Kube-scheduler leader election behavior has been restored for 1.9 and 1.10.
• Kubeadm bootstrap certs are no longer signed by the Kubernetes org.
• Set-based selectors now work properly with StatefulSets.
• Kube-proxy now correctly handles IPv4 vs IPv6 chains in iptables.

Graph of the Week

Josh Berkus worked with Lukasz of the CNCF to add a chart in order to compare workloads across releases. This shows the number of open issues and PRs by day for a user-selected period, allowing us to compare v1.10 with v1.9 and v1.8 at Code Freeze. While the counts are slightly high due to a Github data problem, this does show that v1.10 is in good shape compared with prior releases. However, part of that is because of fewer commits overall to 1.10, so that’s not entirely a good thing.

Last Week In Kubernetes Development (LWKD) is a product of multiple contributors participating in Kubernetes SIG Contributor Experience. All original content is licensed Creative Commons Share-Alike, although linked content and images may be differently licensed. LWKD does collect some information on readers, see our privacy notice for details.

You may contribute to LWKD by submitting pull requests or issues on the LWKD github repo.