By John Belamaric. Originally posted on CoreDNS.io
We are pleased to announce the release of CoreDNS-1.0.0!
Release 1.0.0 and other recent releases have focused on improving the performance and functionality of the kubernetes plugin, since CoreDNS is now on track to eventually replace kube-dns as the default cluster DNS in Kubernetes.
As part of the Kubernetes proposal, we have shown that CoreDNS not only provides more functionality than kube-dns, but performs much better while using less memory. In our tests,CoreDNS running against a cluster with 5000 services was able to process 18,000 queries per second using 73MB of RAM, while kube-dns achieved 7,000qps using 97MB of RAM. This can be partial ascribed to CoreDNS simpler runtime – a single process instead of a combination of several processes.
CoreDNS also implements a number of Kubernetes-related features that are not part of kube-dns, including:
- Filtering of records by namespace
- Filtering of records by label selector
pods verifiedmode, which ensures that a Pod exists before returning an answer for a
endpoint_pod_nameswhich uses Pod names for service endpoint records if the hostname is not set
autopathwhich provides a server-side implementation of the namespace-specific search path. This can cut down the query latency from pods dramatically.
As a general-purpose DNS server, CoreDNS also enables many other use cases that would be difficult or impossible to achieve with kube-dns, such as the ability to create custom DNS entries.
We are excited to continue our contributions to the Kubernetes community, and CoreDNS is being incorporated as a 1.9 alpha feature into a variety of Kubernetes deployment mechanisms, including upcoming versions of kubeadm, kops, minikube, and kubespray.
Of course, there is more to 1.0.0 than just the Kubernetes work. See below for the details on all the changes.
- Fixed a bug in the gRPC server that prevented dnstap from working with it.
- Additional fuzz testing to ferret out obscure bugs.
- Documentation and configuration cleanups.
- log no longer accepts
stdoutin the configuration (use of a file was removed in a previous release). All logging is always to STDOUT. This is a backwards incompatible change, so be sure to check your Corefile for this.
- health now checks plugins that support it for health and reflects that in the server health.
- kubernetes now shows healthy only after the initial API sync is complete.
- kubernetes has bug fixes and performance improvements.
- kubernetes now has an option to use pod names instead of IPs in service endpoint records when the
hostnameis not set.
- metrics have been revised to provide better histograms. You will need to change your Prometheus queries as metric names have changed to comply with Prometheus best practices.
- erratic now supports the health check.
The following people helped with getting this release done: Andy Goldstein, Ben Kochie, Brian Akins, Chris O’Haver, Christian Nilsson, John Belamaric, Max Schmitt, Michael Grosser, Miek Gieben, Ruslan Drozhdzh, Uladzimir Trehubenka, Yong Tang.
If you want to help, please check out one of the issues and start coding!
For documentation and help, see our community page.
Latest posts by Natasha Woods (see all)
- Cloud Native Computing Foundation Announces Bloomberg as Winner of Top End User Award - May 2, 2018
- Cloud Native Computing Foundation Announces DigitalOcean as Gold Member - May 2, 2018
- Cloud Native Computing Foundation Announces Launch of Certified Kubernetes Application Developer (CKAD) Exam - May 2, 2018