The Sonobuoy project automates end-to-end conformance testing for your Kubernetes stack. It also extracts detailed information about your current cluster you can use to monitor cluster changes or catch up with in-house development.
If you need help from the CNCF, for example for switching emails for your CLA signature, the support page has moved.
Steering Committee Update
The Steering Committee discussed potentially getting a community-owned paid GSuite account in order to give community moderators access to Google support for document and mailing list moderation, as well as assinging kubernetes-based email aliases. Next step is discussing this with the CNCF.
All Cloud Provider SIGs have agreed to merge under a single SIG-Cloud-Provider, ideally before KubeCon Barcelona. Also, the SC discussed the problem that the Kubernetes Blog is currently not owned by any Kubernetes SIG, and suggested that SIG-Docs should own it.
We have begun the 1.15 release cycle! Thanks to Test-Infra improvements, we chewed through the 1.15 PR backlog in 24 hours after Code Freeze ended.
dep and Bazel are still deeply entrenched in our ecosystem, Go 1.11 modules are starting to get more traction. This PR adds some basic steps to follow to use client-go as a dependency through the modules system. This is far from final; if you’re going to use this in anything serious, it would be a good idea to throw a watch on #74877 which will follow up with much more goodness. Steps towards a
The ServiceAccount controller has been switched over to the standard TokenRequest API rather than the old, homegrown system. This should improve standardization for token management. In the long term, the token controller will be dropped.
For anyone working on CRI implementations or related code, there will now be a
[k8s.io/cri-api](http://k8s.io/cri-api) library to import from. As with all staging repos, this is replicated automatically from the
k/k repo, but you can get faster compile times and smaller binaries by using it.
More relevant to other projects and teams deploying Prow themselves than for contributors to
k/k itself, Tide will now support a per-PR override for the merge type used. You can specify that a given PR should be squashed or merged with individual commits, no matter what the project default is.
If you’re involved with any of the cloud SIGs, hopefully you’ve already seen this but for the rest of us: change is in the cloudy air! The various cloud provider SIGs are unifying under a shared SIG Cloud Provider banner to better align their efforts and hopefully improve cooperation between the major vendors.
And finally a simple but happy-making PR, our basic documentation is now available in German! Our community gets bigger every day and with that comes more translations. Awesome work by the website team getting this initial batch of translations done.
Because of timing, these include both new 1.15 features and last-minute fixes to 1.14 (or even earlier), and fixes that will get rolled into 1.14.1. We’re omitting the many cherry-picks.
- work around the proxy problems in go 1.12 until we can replace it
- scheduler config now supports YAML as well as JSON
- decrease verbosity of openapi aggregation logs to prevent flooding the logs due to 1.14’s more frequent polling
- job status is now included in the cronjob event for easier success tracking
- Kubelet now asks for protobufs where they are available from the APIserver
- broad refactor of the AWS credential provider for 1.15
- fix vSphere SAML
- get kubectl to display inline CSI volumes now that those are in 1.14
- no more smart labels in
kubectl describefor fields with special characters
- client-go dynamic methods won’t panic on an empty Name
- OpenStack CP can now use Secrets for user credentials
get-kube-binaries.shto fetch a client for a different OS/Arch
- many, many fixes for the quota controller, including using a dynamic informer
- the cloud-controller-manager now uses per-controller configs
- unbound PVCs now count towards the attach limit
- allow enabling watch cache for Events
- support additional taint formats in the client
There were many of these in the last two weeks:
upgrade plannow defaults to stable versions of Kubernetes
- Allow users to migrate from KubeDNS to CoreDNS with the same config
join control-planenow supports external etcd
- delete multiple bootstrap tokens
- no more v1alpha3
- overhaul the IPVS check
- The podTimestamp func is gone.