We’re creating WG-Naming to resolve some longstanding naming problems in Kubernetes, mainly offensive/misleading terms like “whitelist”. If you’re really good with grep and sed, contact Stephen about helping out.
SIG-Contribex has a marketing team who can help you if your SIG needs to recruit more contributors.
gcr.io/google-containers started the flip to
gcr.io/k8s-artifacts-prod on Monday. Everything should be changed by Thursday, hopefully.
Next Deadline: Finish Docs PRs, July 6th
Please spend this week working on your documentation for any 1.19 features. Remember, no docs means no feature.
Also: Code Freeze is July 9th.
The Ingress system has graduated to to GA and
networking.k8s.io/v1! This version bump does come with a number of schema and validation changes, most notably that
defaultBackend is now more clearly labeled as such and
pathType is a required value. Not all Ingress Controllers even support the new path type field but now is a good time to look over your Ingress usage and start planning which modes will work best for you. This was joined by a kubectl PR to add v1 output formatting and improved support for the largely-nascent resource backend system.
Building on the modular formatting system from last month, we now have bona fide JSON logging output. As the structured logging system is used in more places, the context fields will improve, but already you can use this with an existing JSON logging pipeline or analysis system. As you add new logging output, make sure to check out guide to using the structured logging API and add context fields!
While the WG-Naming group is still getting formed, some work to improve outdated terminology has begun. If you are interested in learning more about the working group, check out the mailing list or, once they begin, join in on the meetings.
- API Priority/Fairness (APF) has debugging endpoints
- Pods which have not started are not ready until they start
- kube-proxy sets IP family based on nodeIP when there is no bind address (instead of always IPv4)
- kubectl help has docs for
- Keep deeply nested
kubectl applyfrom taking forever
- Eliminate double-NATing VXLAN in kube-proxy and the second to minute delay in some connections
- Azure uses force detach for volumes
- Log lines about images include the namespace (and
--namespacesin clusterdump works now)
- Since UDP port forwarding was never implemented, maybe warn users about it
- NodePort allocations rely on etcd, avoiding multimaster race conditions
- aggregator_unavailable_apiservice_count metric is now aggregator_unavailable_apiservice_total
- Variable HugePage sizes to Beta and on by default
kubectl get --exportflag, deprecated in 1.14, removed
- golang to v1.14.4 in v1.19
- go.etcd.io/bbolt to v1.3.5 in 1.19
- golang.org/x/text/encoding/unicode to v0.3.3 to patch a security hole; likely to be backported
Original Source: http://lwkd.info/2020/20200623