v0.9.0-alpha.0 (2021-02-02)
Chore
- bump dependencies (via dependabot)
- fix import path for fsnotify
- add dependabot config
- enable virtio-balloon and monitor in QEMU provisioner
- update protobuf, grpc-go, prototool
- update upgrade test version used
Docs
- update components.md
- add v0.9 docs
- add modes to validate command
- document omitting DiskPartition size
- update references to 0.8.0, add 0.8.0 AWS AMIs
- fix latest docs
- set latest docs to v0.8
- provide AMIs for 0.8.0-beta.0
- fix SBC docs to point to beta.0 instead of beta.1
- update Talos release for SBCs
Feat
- move to ECDSA keys for all Kubernetes/etcd certs and keys
- update kernel
- mount hugetlbfs
- allow fqdn to be used when registering k8s node
- copy cryptsetup executable from pkgs
- use multi-arch images for k8s and Flannel CNI
- replace bootkube with Talos-managed control plane
- implement resource API in Talos
- update Linux to 5.10.7, musl-libc to 1.2.2
- update Kubernetes to 1.20.2
- support Wireguard networking
- bump pkgs for kernel with CONFIG_IPV6_MULTIPLE_TABLES
- support type filter in list API and CLI
- add commands to manage/query etcd cluster
- support disk image in talosctl cluster create
- update Kubernetes to 1.20.1
Fix
- use hugetlbfs instead of none
- use grpc load-balancing when connecting to trustd
- lower memory usage a bit by disabling memory profiling
- don't probe disks in container mode
- prefix rendered Talos-owned static pod manifests
- bump timeout for worker apid waiting for kubelet client config
- kill all processes and umount all disk on reboot/shutdown
- open blockdevices with exclusive flock for partitioning
- list command unlimited recursion default behavior
- pick first interface valid hostname (vs. last one)
- allow 'console' argument in kernel args to be always overridden
- bring up bonded interfaces correctly on packet
- checkpoint controller-manager and scheduler
- correctly transport gRPC errors from apid
- use SetAll instead of AppendAll when building kernel args
- add more dependencies for bootstrap services
- pass disk image flags to e2e-qemu cluster create command
- ignore pods spun up from checkpoints in health checks
- leave etcd for staged upgrades
- ignore errors on stopping/removing pod sandboxes
- use the correct console on Banana Pi M64
- don't run LabelNodeAsMaster in two sequences
Refactor
- update go-blockdevice and restructure disk interaction code
- define default kernel flags in machinery instead of procfs
Release
- v0.9.0-alpha.0: prepare release
Test
- clear connection refused errors after reset
- skip etcd tests on non-HA clusters
Images
quay.io/coreos/flannel:v0.13.0
ghcr.io/talos-systems/install-cni:v0.1.0-6-gdc32cc8
k8s.gcr.io/coredns:1.7.0
gcr.io/etcd-development/etcd:v3.4.14
k8s.gcr.io/kube-apiserver:v1.20.2
k8s.gcr.io/kube-controller-manager:v1.20.2
k8s.gcr.io/kube-scheduler:v1.20.2
k8s.gcr.io/kube-proxy:v1.20.2
ghcr.io/talos-systems/kubelet:v1.20.2
ghcr.io/talos-systems/installer:v0.9.0-alpha.0
k8s.gcr.io/pause:3.2