diff --git a/.ci/podman.mk b/.ci/podman.mk index 626ad4f6..fa905a57 100644 --- a/.ci/podman.mk +++ b/.ci/podman.mk @@ -38,7 +38,8 @@ clean: rm-test-image rm-image .PHONY: rm-remote-untagged rm-remote-untagged: @echo "Removing all untagged images from $(IMAGE) in $(REGION)" - @aws ecr-public batch-delete-image --repository-name $(IMAGE) --region $(REGION) --image-ids $$(for image in $$(aws ecr-public describe-images --repository-name $(IMAGE) --region $(REGION) --output json | jq -r '.imageDetails[] | select(.imageTags | not ).imageDigest'); do echo -n "imageDigest=$$image "; done) + @IMAGE_IDS=$$(for image in $$(aws ecr-public describe-images --repository-name $(IMAGE) --region $(REGION) --output json | jq -r '.imageDetails[] | select(.imageTags | not ).imageDigest'); do echo -n "imageDigest=$$image "; done) ; \ + [ -n "$$IMAGE_IDS" ] && aws ecr-public batch-delete-image --repository-name $(IMAGE) --region $(REGION) --image-ids $$IMAGE_IDS || echo "Nothing to remove" .PHONY: rm-image rm-image: diff --git a/admin/libhelm.sh b/admin/libhelm.sh index 56d57023..cc2291dc 100644 --- a/admin/libhelm.sh +++ b/admin/libhelm.sh @@ -137,6 +137,9 @@ function _helm() { local action=$1 local module=$2 + # check if module is even enabled and return if not + [ ! -f $WORKDIR/kubezero/templates/${module}.yaml ] && { echo "Module $module disabled. No-op."; return 0; } + local chart="$(yq eval '.spec.source.chart' $WORKDIR/kubezero/templates/${module}.yaml)" local namespace="$(yq eval '.spec.destination.namespace' $WORKDIR/kubezero/templates/${module}.yaml)" diff --git a/charts/kubezero-sql/Chart.yaml b/charts/kubezero-sql/Chart.yaml index 54026107..4b4a7860 100644 --- a/charts/kubezero-sql/Chart.yaml +++ b/charts/kubezero-sql/Chart.yaml @@ -2,7 +2,7 @@ apiVersion: v2 name: kubezero-sql description: KubeZero umbrella chart for SQL databases like MariaDB, PostgreSQL type: application -version: 0.2.0 +version: 0.2.1 home: https://kubezero.com icon: https://cdn.zero-downtime.net/assets/kubezero/logo-small-64.png keywords: @@ -17,7 +17,7 @@ dependencies: version: ">= 0.1.5" repository: https://cdn.zero-downtime.net/charts/ - name: mariadb-galera - version: 7.4.2 + version: 7.4.3 repository: https://charts.bitnami.com/bitnami condition: mariadb-galera.enabled kubeVersion: ">= 1.20.0" diff --git a/charts/kubezero-sql/values.yaml b/charts/kubezero-sql/values.yaml index 9262af14..768a8891 100644 --- a/charts/kubezero-sql/values.yaml +++ b/charts/kubezero-sql/values.yaml @@ -1,5 +1,5 @@ mariadb-galera: - enabled: true + enabled: false replicaCount: 2 diff --git a/docs/v1.23.md b/docs/v1.23.md index 2fb2bef6..0b5d2399 100644 --- a/docs/v1.23.md +++ b/docs/v1.23.md @@ -2,115 +2,41 @@ ## What's new - Major themes -- update inf1 neuron drivers incl. node auto-taints -- support for Nvidia g5 instances incl. the whole toolchain up to device drivers etc, auto node taints -- ExtendedResourceToleration AdmissionController enabled to auto tolerate INF1 and Nvidia pods -- Cluster-Autoscaler - - - -### Alpine - Custom AMIs -Starting with 1.22, all KubeZero nodes boot using custom AMIs. These AMIs will be provided and shared by the Zero Down Time for all customers. As always, all sources incl. the build pipeline are freely available [here](https://git.zero-downtime.net/ZeroDownTime/alpine-zdt-images). - -This eliminates *ALL* dependencies at boot time other than container registries. Gone are the days when Ubuntu, SuSE or Github decided to ruin your morning coffee. - -KubeZero migrates from Ubuntu 20.04 LTS to [Alpine v3.15](https://www.alpinelinux.org/releases/) as its base OS. -#### Highlights: -- minimal attack surface by removing all unnecessary bloat, -like all things SystemD, Ubuntu's snap, etc -- reduced root file system size from 8GB to 2GB -- minimal memory consumption of about 12MB fully booted - - *Minimal* fully booted instance incl. SSH and Monit: - - | | Ubuntu | Alpine| - |-|--------|-----| - | Memory used | 60MB | 12 MB | - | RootFS used | 1.1GB | 330 MB | - | RootFS encrypted | no | yes | - | Kernel | 5.11 | 5.15 | - | Init | Systemd | OpenRC | - | AMI / EBS size | 8GB | 1GB | - | Boot time | ~120s | ~45s | - -- Encrypted AMIs: -This closes the last gaps you might have in achieving *full encryption at rest* for every volume within a default KubeZero deployment. - -### Etcd -On AWS a new dedicated GP3 EBS volume gets provisioned per controller and is used as persistent etcd storage. These volumes will persist for the life time of the cluster and reused by future controller nodes in each AZ. -This ensure no data loss during upgrade or restore situations of single controller clusters. The hourly backup on S3 will still be used as fallback / disaster recovery option in case the file system gets corrupted etc. - - -### DNS -The [external-dns](https://github.com/kubernetes-sigs/external-dns) controller got integrated and is used to provide DNS based loadbalacing for the apiserver itself. This allows high available control planes on AWS as well as bare-metal in combination with various DNS providers. - -Further usage of this controller to automate any DNS related configurations, like Ingress etc. is planned for following releases. - -### Container runtime -Cri-o now uses crun rather than runc, which reduces the memory overhead *per pod* from 16M to 4M, details at [crun intro](https://www.redhat.com/sysadmin/introduction-crun) - -With 1.22 and the switch to crun, support for [CgroupV2](https://www.kernel.org/doc/Documentation/cgroup-v2.txt) has been enabled. - -### AWS Neuron INF support -Initial support for [Inf1 instances](https://aws.amazon.com/ec2/instance-types/inf1/) part of [AWS Neuron](https://aws.amazon.com/machine-learning/neuron/). - -Workers automatically load the custom kernel module on these instance types and expose the `/dev/neuron*` devices. +- Cilium added as second CNI to prepare full migration to Cilium with 1.24 upgrade +- support for Nvidia g5 instances incl. pre-installed kernel drivers, cudo toolchain and CRI intergration +- updated inf1 neuron drivers +- ExtendedResourceToleration AdmissionController and auto-taints allowing Neuron and Nvidia pods ONLY to be scheduled on dedicated workers +- full Cluster-Autoscaler integration ## Version upgrades -- Istio to 1.13.3 using the new Helm [gateway charts](https://istio.io/latest/docs/setup/additional-setup/gateway/) -- Logging: ECK operator upgraded from 1.6 to 2.1, fluent-bit 1.9.3 -- Metrics: Prometheus and all Grafana charts to latest to match V1.22 -- ArgoCD to V2.2.5 +- Istio to 1.14.4 +- Logging: ECK operator to 2.4, fluent-bit 1.9.8 +- Metrics: Prometheus and all Grafana charts to latest to match V1.23 +- ArgoCD to V2.4 ( access to pod via shell disabled by default ) - AWS EBS/EFS CSI drivers to latest versions -- cert-manager to V1.8 -- aws-termination-handler to 1.16 -- aws-iam-authenticator to 0.5.7, required for >1.22 which allows using the latest version on the client side again - -## Misc -- new metrics and dashboards for openEBS LVM CSI drivers -- new node label `node.kubernetes.io/instance-type` for all nodes containing the EC2 instance type -- kubelet root moved to `/var/lib/containers` to ensure ephemeral storage is allocated from the configurable volume rather than the root fs of the worker - +- cert-manager to V1.9.1 # Upgrade `(No, really, you MUST read this before you upgrade)` - Ensure your Kube context points to the correct cluster ! -- Ensure any usage of Kiam has been migrated to OIDC providers as any remaining Kiam components will be deleted as part of the upgrade -1. Migrate ArgoCD KubeZero config: - `cat | ./releases/v1.22/migrate_agro.py` and adjust if needed and replace the original. Do NOT commit yet ! +1. Enable `containerProxy` for NAT instances and upgrade NAT instance using the new V2 Pulumi stacks -2. Upgrade `logging` and `metrics` module -- `kubectl get crd elasticsearches.elasticsearch.k8s.elastic.co && kubectl replace -f https://download.elastic.co/downloads/eck/2.1.0/crds.yaml` CRDs for logging -- `./bootstrap.sh apply logging ` logging module to support new OS coming with 1.22 -- `./bootstrap.sh crds metrics ` CRDs for metrics -- `./bootstrap.sh apply metrics ` to get new exporters in place to support 1.22 +2. Review CFN config for controller and workers ( enable containerProxy, remove legacy version settings etc ) -3. Trigger the cluster upgrade: -`./release/v1.22/upgrade_cluster.sh` +3. Upgrade CFN stacks for the control plane and all worker groups -4. Upgrade CFN stacks for the control plane and all worker groups -Change Kubernetes version in controller config from `1.21.9` to `1.22.8` +4. Trigger fully-automated cluster upgrade: + `./admin/upgrade_cluster.sh ` -5. Reboot controller(s) one by one +5. Reboot controller(s) one by one Wait each time for controller to join and all pods running. Might take a while ... -6. Launch new set of workers, at least enough to host new Istio Ingress gateways due to Kernel requirements -Eg. by doubling `desired` for each worker ASG, +6. Launch new set of workers eg. by doubling `desired` for each worker ASG + once new workers are ready, cordon and drain all old workers +The cluster-autoscaler will remove the old workers automatically after about 10min ! -7. Upgrade via boostrap.sh -As the changes around Istio are substantial in this release we need to upgrade some parts step by step to prevent service outages, especially for private-ingress. - -- `./bootstrap.sh crds all ` to deploy all new CRDs first -- `./bootstrap.sh apply cert-manager ` to update cert-manager, required for Istio -- `./bootstrap.sh apply istio ` to update the Istio control plane -- `./bootstrap.sh apply istio-private-ingress ` to deploy the new private-ingress gateways first -- `./bootstrap.sh apply istio-ingress ` to update the public ingress and also remove the 1.21 private-ingress gateways - -8. Finalize via ArgoCD - git add / commit / pusSh `` and watch ArgoCD do its work. - -9. Drain old workers - Drain one by one and reset each ASG to initial "desired" value. +7. If all looks good, commit the ArgoApp resouce for Kubezero, before re-enabling ArgoCD itself. + git add / commit / push ``