cozystack/packages/apps/kubernetes

Managed Kubernetes Service

Managed Kubernetes in Cozystack

Whenever you want to deploy a custom containerized application in Cozystack, it's best to deploy it to a managed Kubernetes cluster.

Cozystack deploys and manages Kubernetes-as-a-service as standalone applications within each tenant’s isolated environment. In Cozystack, such clusters are named tenant Kubernetes clusters, while the base Cozystack cluster is called a management or root cluster. Tenant clusters are fully separated from the management cluster and are intended for deploying tenant-specific or customer-developed applications.

Within a tenant cluster, users can take advantage of LoadBalancer services and easily provision physical volumes as needed.
The control-plane operates within containers, while the worker nodes are deployed as virtual machines, all seamlessly managed by the application.

Kubernetes version in tenant clusters is independent of Kubernetes in the management cluster. Users can select the supported patch versions from 1.31 to 1.35.

Why Use a Managed Kubernetes Cluster?

Kubernetes has emerged as the industry standard, providing a unified and accessible API, primarily utilizing YAML for configuration. This means that teams can easily understand and work with Kubernetes, streamlining infrastructure management.

Kubernetes leverages robust software design patterns, enabling continuous recovery in any scenario through the reconciliation method. Additionally, it ensures seamless scaling across a multitude of servers, addressing the challenges posed by complex and outdated APIs found in traditional virtualization platforms. This managed service eliminates the need for developing custom solutions or modifying source code, saving valuable time and effort.

The Managed Kubernetes Service in Cozystack offers a streamlined solution for efficiently managing server workloads.

Starting Work

Once the tenant Kubernetes cluster is ready, you can get a kubeconfig file to work with it. It can be done via UI or a kubectl request:

• Open the Cozystack dashboard, switch to your tenant, find and open the application page. Copy one of the config files from the Secrets section.

• Run the following command (using the management cluster kubeconfig):

  kubectl get secret -n tenant-<name> kubernetes-<clusterName>-admin-kubeconfig -o go-template='{{ printf "%s\n" (index .data "admin.conf" | base64decode) }}' > admin.conf
  

There are several kubeconfig options available:

• admin.conf — The standard kubeconfig for accessing your new cluster. You can create additional Kubernetes users using this configuration.
• admin.svc — Same token as admin.conf, but with the API server address set to the internal service name. Use it for applications running inside the cluster that need API access.
• super-admin.conf — Similar to admin.conf, but with extended administrative permissions. Intended for troubleshooting and cluster maintenance tasks.
• super-admin.svc — Same as super-admin.conf, but pointing to the internal API server address.

Implementation Details

A tenant Kubernetes cluster in Cozystack is essentially Kubernetes-in-Kubernetes. Deploying it involves the following components:

• Kamaji Control Plane: Kamaji is an open-source project that facilitates the deployment of Kubernetes control planes as pods within a root cluster. Each control plane pod includes essential components like kube-apiserver, controller-manager, and scheduler, allowing for efficient multi-tenancy and resource utilization.

• Etcd Cluster: A dedicated etcd cluster is deployed using Ænix's etcd-operator. It provides reliable and scalable key-value storage for the Kubernetes control plane.

• Worker Nodes: Virtual Machines are provisioned to serve as worker nodes using KubeVirt. These nodes are configured to join the tenant Kubernetes cluster, enabling the deployment and management of workloads. Worker node disks automatically detect and match the underlying volume's block size (blockSize.matchVolume) to ensure compatibility with storage backends that use non-512-byte sectors, such as LINSTOR DRBD with 4Kn drives.

• Cluster API: Cozystack is using the Kubernetes Cluster API to provision the components of a cluster.

This architecture ensures isolated, scalable, and efficient tenant Kubernetes environments.

See the reference for components utilized in this service:

• Kamaji Control Plane
• Kamaji — Cluster API
• github.com/clastix/kamaji
• KubeVirt
• github.com/kubevirt/kubevirt
• github.com/aenix-io/etcd-operator
• Kubernetes Cluster API
• github.com/kubernetes-sigs/cluster-api-provider-kubevirt
• github.com/kubevirt/csi-driver

Breaking Changes

• ephemeralStorage renamed to diskSize (v1.4): The nodeGroups[name].ephemeralStorage field has been renamed to nodeGroups[name].diskSize. Existing clusters are migrated transparently by platform migration 41 during the pre-upgrade hook — no manual action is required. Newly written values should use diskSize. Existing VMs are rolling-updated via CAPI when the new values are applied. With the Talos worker rollover in this release the field now sizes the single system disk (Talos OS image streamed from factory.talos.dev, kubelet state, containerd image cache, local-path PVCs) — the pre-Talos disk-kubelet PVC layout was removed. State on that single disk persists across same-VM reboots (virt-launcher restart, guest reboot, node failure); VM replacement by CAPI (e.g. nodeGroup field change, MachineHealthCheck remediation) provisions a fresh disk.

• Air-gapped tenant workers temporarily unsupported (Phase 1 Talos rollover): tenant worker bootstrap moves from Ubuntu containerDisk + kubeadm to a Talos image fetched over HTTPS from factory.talos.dev and a factory.talos.dev/installer/... installer reference, neither of which is currently overridable. The Helm-rendered *-patch-containerd Secret that previously plumbed the platform-wide registries.mirrors config to tenant workers (via the kubeadm template's containerd certs.d mount) has no consumer in the Talos machineconfig and was removed in this release. Tenants behind a strict-egress firewall or relying on registries.mirrors for tenant workers will be unable to create new workers until Phase 2 restores this via machine.registries.mirrors knobs and an installable/imageBase override. Tracked as a follow-up to this Phase 1 PR — file an issue if you depend on this and it is not yet landed.

• Worker MachineHealthCheck remediation is now enabled by default (Phase 1 Talos rollover): the worker MHC maxUnhealthy moved from a hard-coded 0 (remediation effectively off) to a configurable nodeHealthCheck.maxUnhealthy defaulting to "50%". CAPI now auto-remediates (deletes and replaces) unhealthy worker Machines while at least 50% of a group stays healthy. "50%" deliberately leaves headroom for transient unhealthy nodes during the kubeadm→Talos rollover and slow first boots from factory.talos.dev; set nodeHealthCheck.maxUnhealthy: "0%" to keep the previous remediation-off behaviour until your fleet is stable on Talos workers.

• GitOps-managed tenant Kubernetes CRs must bump spec.version in Git (Phase 1 Talos rollover): platform migration 46 patches live kuberneteses.apps.cozystack.io objects from v1.30 to v1.31 (v1.30 left the Talos↔Kubernetes support matrix). When the tenant Kubernetes CR is reconciled from Git (Flux/Argo), the next source reconcile re-applies version: v1.30 over migration 46's patch, which then trips the chart's _versions.tpl guard and the HelmRelease fails. Update spec.version to v1.31 (or newer) in your Git source before/with the platform upgrade. Tenants managed via the API or dashboard need no action — migration 46 handles them.

• Remote-accessible LINSTOR StorageClasses are auto-propagated to tenants (v1.5): infra-cluster LINSTOR StorageClasses whose linstor.csi.linbit.com/allowRemoteVolumeAccess is not "false" are created inside each tenant under the same name (provisioned by csi.kubevirt.io). Upgrade caveat: delete any manually created tenant StorageClass whose name collides with a propagated infra class (e.g. a hand-made replicated) before upgrading, or the propagated class cannot be created and the tenant CSI release will not converge. Infra classes that must stay node-local need allowRemoteVolumeAccess: "false" set explicitly — an absent annotation is treated as remote-accessible. Propagation is evaluated at HelmRelease render time, so classes added or removed on the infra cluster after a tenant exists propagate only on that tenant's next reconcile.

The top-level storageClass field is annotated as immutable in the chart schema — see docs/storage-immutability.md for the contract and which consumers enforce it. The per-node-group nodeGroups[name].storageClass field is intentionally not annotated immutable: it is optional and undefaulted, so a strict self == oldSelf rule would block any future attempt to set it on an existing node group.

Parameters

Common Parameters

Name	Description	Type	Value
storageClass	Default StorageClass inside the tenant cluster. Remote-accessible LINSTOR classes are auto-propagated to the tenant under the same name.	string	replicated

Application-specific Parameters

Name	Description	Type	Value
nodeGroups	Worker nodes configuration map. When left empty, a single default node group md0 (one ingress-nginx worker) is provisioned. Provide your own groups to take full control — they are not merged with the default, so you may name and omit groups freely.	map[string]object	{}
nodeGroups[name].minReplicas	Minimum number of replicas.	int	0
nodeGroups[name].maxReplicas	Maximum number of replicas.	int	10
nodeGroups[name].instanceType	Virtual machine instance type.	string	u1.medium
nodeGroups[name].diskSize	System disk size for the worker VM. Carries the Talos OS image (factory.talos.dev raw artifact streamed in by CDI), kubelet state, containerd image cache, and any local-path PVCs. Pre-Talos installs used a separate disk-kubelet PVC for kubelet/containerd state; on Talos this is consolidated onto the single system disk imaged from the factory artifact.	quantity	20Gi
nodeGroups[name].storageClass	StorageClass for worker node persistent disks. When empty, uses the management cluster default StorageClass (the one annotated storageclass.kubernetes.io/is-default-class: true). NOTE: deliberately not marked immutable — the field is optional and undefaulted, so a strict self == oldSelf rule would block any future attempt to set it on an existing node group.	string	""
nodeGroups[name].roles	List of node roles.	[]string	[]
nodeGroups[name].resources	Explicit CPU and memory for each worker node, as an alternative to instanceType sizing. Optional: when omitted, the node is sized by instanceType. When both cpu and memory are set, they take precedence and instanceType is ignored for that node group (the instancetype is omitted from the VM, since KubeVirt cannot override an instancetype's CPU/memory). Set both cpu and memory together or neither; setting only one is rejected at render time.	object	{}
nodeGroups[name].resources.cpu	CPU available.	quantity	""
nodeGroups[name].resources.memory	Memory (RAM) available.	quantity	""
nodeGroups[name].gpus	List of GPUs to attach (NVIDIA driver requires at least 4 GiB RAM).	[]object	[]
nodeGroups[name].gpus[i].name	Name of GPU, such as "nvidia.com/AD102GL_L40S".	string	""
nodeGroups[name].kubelet	Kubelet resource reservations for this node group.	object	{}
nodeGroups[name].kubelet.systemReservedMemory	Memory reserved for host OS. Auto-computed from instanceType if empty.	string	""
nodeGroups[name].kubelet.kubeReservedMemory	Memory reserved for kubelet and container runtime. Auto-computed from instanceType if empty.	string	""
nodeGroups[name].kubelet.systemReservedCpu	CPU reserved for host OS. Auto-computed from instanceType if empty.	string	""
nodeGroups[name].kubelet.kubeReservedCpu	CPU reserved for kubelet and container runtime. Auto-computed from instanceType if empty.	string	""
nodeGroups[name].kubelet.evictionHardMemory	Hard eviction threshold for memory (absolute like 200Mi or percentage like 7%).	string	7%
nodeGroups[name].kubelet.evictionSoftMemory	Soft eviction threshold for memory (absolute like 1Gi or percentage like 10%).	string	10%
version	Kubernetes major.minor version to deploy	string	v1.35
host	External hostname for Kubernetes cluster. Defaults to <cluster-name>.<tenant-host> if empty.	string	""

Cluster Addons

Name	Description	Type	Value
addons	Cluster addons configuration.	object	{}
addons.certManager	Cert-manager addon.	object	{}
addons.certManager.enabled	Enable cert-manager.	bool	false
addons.certManager.valuesOverride	Custom Helm values overrides.	object	{}
addons.cilium	Cilium CNI plugin.	object	{}
addons.cilium.valuesOverride	Custom Helm values overrides.	object	{}
addons.gatewayAPI	Gateway API addon.	object	{}
addons.gatewayAPI.enabled	Enable Gateway API.	bool	false
addons.ingressNginx	Ingress-NGINX controller.	object	{}
addons.ingressNginx.enabled	Enable the controller (requires nodes labeled ingress-nginx).	bool	false
addons.ingressNginx.exposeMethod	Method to expose the controller. Allowed values: Proxied, LoadBalancer.	string	Proxied
addons.ingressNginx.hosts	Domains routed to this tenant cluster when exposeMethod is Proxied.	[]string	[]
addons.ingressNginx.valuesOverride	Custom Helm values overrides.	object	{}
addons.gpuOperator	NVIDIA GPU Operator.	object	{}
addons.gpuOperator.enabled	Enable GPU Operator.	bool	false
addons.gpuOperator.valuesOverride	Custom Helm values overrides.	object	{}
addons.hami	HAMi GPU virtualization middleware.	object	{}
addons.hami.enabled	Enable HAMi (requires GPU Operator).	bool	false
addons.hami.valuesOverride	Custom Helm values overrides.	object	{}
addons.fluxcd	FluxCD GitOps operator.	object	{}
addons.fluxcd.enabled	Enable FluxCD.	bool	false
addons.fluxcd.valuesOverride	Custom Helm values overrides.	object	{}
addons.monitoringAgents	Monitoring agents.	object	{}
addons.monitoringAgents.enabled	Enable monitoring agents.	bool	false
addons.monitoringAgents.valuesOverride	Custom Helm values overrides.	object	{}
addons.verticalPodAutoscaler	Vertical Pod Autoscaler.	object	{}
addons.verticalPodAutoscaler.valuesOverride	Custom Helm values overrides.	object	{}
addons.velero	Velero backup/restore addon.	object	{}
addons.velero.enabled	Enable Velero.	bool	false
addons.velero.valuesOverride	Custom Helm values overrides.	object	{}
addons.coredns	CoreDNS addon.	object	{}
addons.coredns.valuesOverride	Custom Helm values overrides.	object	{}
addons.ouroboros	Hairpin-NAT fix for ingress-nginx with PROXY-protocol.	object	{}
addons.ouroboros.enabled	Enable ouroboros. Requires addons.ingressNginx.enabled (chart-render fail otherwise). Only useful when PROXY-protocol is wired on the tenant ingress-nginx via valuesOverride.	bool	false
addons.ouroboros.valuesOverride	Custom Helm values overrides. Operator-key wins over cozystack defaults.	object	{}

Kubernetes Control Plane Configuration

Name	Description	Type	Value
controlPlane	Kubernetes control-plane configuration.	object	{}
controlPlane.replicas	Number of control-plane replicas.	int	2
controlPlane.apiServer	API Server configuration.	object	{}
controlPlane.apiServer.resources	CPU and memory resources for API Server.	object	{}
controlPlane.apiServer.resources.cpu	CPU available.	quantity	""
controlPlane.apiServer.resources.memory	Memory (RAM) available.	quantity	""
controlPlane.apiServer.resourcesPreset	Preset if resources omitted.	string	c1.medium
controlPlane.apiServer.extraArgs	Extra command-line flags appended to the tenant kube-apiserver, passed through to KamajiControlPlane spec.apiServer.extraArgs. Use for OIDC SSO (--oidc-issuer-url, --oidc-client-id) or structured authentication (--authentication-config=/etc/kubernetes/authentication/config.yaml, --requestheader-uid-headers=X-Remote-Uid). Empty by default (no change to current behavior).	[]string	[]
controlPlane.apiServer.extraVolumes	Extra volumes added to the control-plane Deployment, passed through to KamajiControlPlane spec.deployment.extraVolumes. Use to mount a ConfigMap or Secret holding an AuthenticationConfiguration file referenced by extraArgs. Each item is a core/v1 Volume, but because the control-plane pod runs on the management cluster only configMap and secret sources are allowed (host-reaching sources like hostPath/csi and token-projection via projected are rejected); each volume must have a unique, non-empty name and exactly one source. The names talos-ca and talos-tls-cert are reserved by the chart. Empty by default.	[]object	[]
controlPlane.apiServer.extraVolumeMounts	Extra volume mounts added to the tenant kube-apiserver container, passed through to KamajiControlPlane spec.apiServer.extraVolumeMounts. Each name must reference a volume declared in extraVolumes; the chart-managed talos secret volumes cannot be mounted. Each item is a core/v1 VolumeMount. Empty by default.	[]object	[]
controlPlane.controllerManager	Controller Manager configuration.	object	{}
controlPlane.controllerManager.resources	CPU and memory resources for Controller Manager.	object	{}
controlPlane.controllerManager.resources.cpu	CPU available.	quantity	""
controlPlane.controllerManager.resources.memory	Memory (RAM) available.	quantity	""
controlPlane.controllerManager.resourcesPreset	Preset if resources omitted.	string	t1.micro
controlPlane.scheduler	Scheduler configuration.	object	{}
controlPlane.scheduler.resources	CPU and memory resources for Scheduler.	object	{}
controlPlane.scheduler.resources.cpu	CPU available.	quantity	""
controlPlane.scheduler.resources.memory	Memory (RAM) available.	quantity	""
controlPlane.scheduler.resourcesPreset	Preset if resources omitted.	string	t1.micro
controlPlane.konnectivity	Konnectivity configuration.	object	{}
controlPlane.konnectivity.server	Konnectivity Server configuration.	object	{}
controlPlane.konnectivity.server.resources	CPU and memory resources for Konnectivity.	object	{}
controlPlane.konnectivity.server.resources.cpu	CPU available.	quantity	""
controlPlane.konnectivity.server.resources.memory	Memory (RAM) available.	quantity	""
controlPlane.konnectivity.server.resourcesPreset	Preset if resources omitted.	string	t1.micro
images	Optional image overrides for air-gapped or rate-limited registries.	object	{}
images.waitForKubeconfig	Image used by the wait-for-kubeconfig init container. Empty falls back to images/busybox.tag.	string	""
images.kubectl	Image used by the bootstrap-token tenant Job (kubectl). Empty falls back to images/kubectl.tag.	string	""
images.talosCsrSigner	Image used by the talos-csr-signer sidecar in the Kamaji control plane. Empty falls back to images/talos-csr-signer.tag.	string	""

Talos Worker Image

Name	Description	Type	Value
talos	Talos worker image configuration.	object	{}
talos.version	Talos release used for worker OS image and installer. Must satisfy the chart's Talos<->Kubernetes support matrix against the chosen version.	string	v1.13.0
talos.schematicID	Talos image-factory schematic ID. Defaults to the cozystack-tested vanilla schematic. Operators using custom schematics (system extensions, kernel args) override here.	string	ce4c980550dd2ab1b17bbf2b08801c7eb59418eafe8f279833297925d67c7515

Node Health Check

Name	Description	Type	Value
nodeHealthCheck	MachineHealthCheck tuning for worker node groups.	object	{}
nodeHealthCheck.maxUnhealthy	Maximum number of unhealthy nodes tolerated per node group before remediation is paused. The MHC admission webhook accepts either a bare integer ("0", "1", ...) or a percentage ("0%", "50%"); bare numeric strings are rejected, so the safer default is to express the value as a percentage. Default "50%" leaves headroom for transient unhealthy nodes during the kubeadm-to-Talos rollover and slow first boots from factory.talos.dev. Drop to "0%" once the fleet is stable on Talos workers.	string	50%
nodeHealthCheck.nodeStartupTimeout	Maximum time a Machine is allowed to spend reaching the Ready condition before it is remediated. Raise for slow first boots (Talos image fetch from factory.talos.dev or a busy storage class on the kubevirt-csi PVC populator).	string	10m

Parameter examples and reference

resources and resourcesPreset

resources sets explicit CPU and memory configurations for each replica. When left empty, the preset defined in resourcesPreset is applied.

resources:
  cpu: 4000m
  memory: 4Gi

resourcesPreset sets named CPU and memory configurations for each replica. This setting is ignored if the corresponding resources value is set.

Preset name	CPU	memory
nano	250m	128Mi
micro	500m	256Mi
small	1	512Mi
medium	1	1Gi
large	2	2Gi
xlarge	4	4Gi
2xlarge	8	8Gi

instanceType Resources

The following instanceType resources are provided by Cozystack:

Name	vCPUs	Memory
cx1.2xlarge	8	16Gi
cx1.2xlarge1gi	8	16Gi
cx1.4xlarge	16	32Gi
cx1.4xlarge1gi	16	32Gi
cx1.8xlarge	32	64Gi
cx1.8xlarge1gi	32	64Gi
cx1.large	2	4Gi
cx1.large1gi	2	4Gi
cx1.medium	1	2Gi
cx1.medium1gi	1	2Gi
cx1.xlarge	4	8Gi
cx1.xlarge1gi	4	8Gi
d1.2xlarge	8	32Gi
d1.2xmedium	2	4Gi
d1.4xlarge	16	64Gi
d1.8xlarge	32	128Gi
d1.large	2	8Gi
d1.medium	1	4Gi
d1.micro	1	1Gi
d1.nano	1	512Mi
d1.small	1	2Gi
d1.xlarge	4	16Gi
gn1.2xlarge	8	32Gi
gn1.4xlarge	16	64Gi
gn1.8xlarge	32	128Gi
gn1.xlarge	4	16Gi
m1.2xlarge	8	64Gi
m1.2xlarge1gi	8	64Gi
m1.4xlarge	16	128Gi
m1.4xlarge1gi	16	128Gi
m1.8xlarge	32	256Gi
m1.8xlarge1gi	32	256Gi
m1.large	2	16Gi
m1.large1gi	2	16Gi
m1.xlarge	4	32Gi
m1.xlarge1gi	4	32Gi
n1.2xlarge	16	32Gi
n1.4xlarge	32	64Gi
n1.8xlarge	64	128Gi
n1.large	4	8Gi
n1.medium	4	4Gi
n1.xlarge	8	16Gi
o1.2xlarge	8	32Gi
o1.4xlarge	16	64Gi
o1.8xlarge	32	128Gi
o1.large	2	8Gi
o1.medium	1	4Gi
o1.micro	1	1Gi
o1.nano	1	512Mi
o1.small	1	2Gi
o1.xlarge	4	16Gi
rt1.2xlarge	8	32Gi
rt1.4xlarge	16	64Gi
rt1.8xlarge	32	128Gi
rt1.large	2	8Gi
rt1.medium	1	4Gi
rt1.micro	1	1Gi
rt1.small	1	2Gi
rt1.xlarge	4	16Gi
u1.2xlarge	8	32Gi
u1.2xmedium	2	4Gi
u1.4xlarge	16	64Gi
u1.8xlarge	32	128Gi
u1.large	2	8Gi
u1.medium	1	4Gi
u1.micro	1	1Gi
u1.nano	1	512Mi
u1.small	1	2Gi
u1.xlarge	4	16Gi

Kubelet Resource Reservations

Each node group supports a kubelet object that configures how much memory and CPU the kubelet reserves for the host OS and the Kubernetes/system components running on the worker node.

When systemReservedMemory or kubeReservedMemory are left empty, they are auto-computed using the following formula:

1. Determine the effective memory of the node:
   • If resources.memory is explicitly set, use that value.
   • Otherwise, look up the instanceType and use its memory.guest value.
   • If neither is available, the reservation falls back to the minimum (256Mi).
2. Calculate 5% of the effective memory (in MiB, rounded down).
3. Clamp the result to the range [256Mi, 1Gi]:
   • Nodes with 5 GiB or less get the minimum 256Mi reservation.
   • Nodes with 20 GiB or more get the maximum 1Gi reservation.

Both systemReservedMemory and kubeReservedMemory receive the same auto-computed value by default.

CPU reservations (systemReservedCpu, kubeReservedCpu) follow the same pattern: 5% of effective CPU, clamped to [50m, 500m]. Both are auto-computed when left empty.

Kubelet Defaults

Parameter	Default	Description
systemReservedMemory	auto-computed	Memory reserved for host OS
kubeReservedMemory	auto-computed	Memory reserved for kubelet and container runtime
systemReservedCpu	auto-computed	CPU reserved for host OS
kubeReservedCpu	auto-computed	CPU reserved for kubelet and container runtime
evictionHardMemory	7%	Hard eviction threshold for memory
evictionSoftMemory	10%	Soft eviction threshold for memory
evictionSoftGracePeriod	1m30s (hardcoded)	Duration a soft eviction threshold must be breached before triggering eviction
evictionMinimumReclaim	256Mi (hardcoded)	Minimum amount of memory reclaimed per eviction action

Eviction thresholds can be specified as percentages (e.g., 7%) or absolute values (e.g., 200Mi). Both thresholds must use the same unit type. The hard threshold must be strictly less than the soft threshold.

The evictionSoftGracePeriod and evictionMinimumReclaim parameters are currently hardcoded in the chart template and cannot be overridden through values.

Capacity Annotation

When kubelet resource reservations are configured, both the capacity.cluster-autoscaler.kubernetes.io/memory and capacity.cluster-autoscaler.kubernetes.io/cpu annotations on MachineDeployments report allocatable values instead of total node resources. Memory allocatable subtracts system-reserved, kube-reserved, and eviction-hard. CPU allocatable subtracts system-reserved and kube-reserved. This aligns the annotations with the values the cluster autoscaler uses in its scaling calculations.

Upgrading from a version without this feature may cause the autoscaler to see reduced per-node capacity after the annotations are updated, which can trigger additional scale-up operations. No action is typically required — the new values reflect the actual memory available for workload scheduling.

Note: When neither resources.memory nor instanceType is set, eviction thresholds (default 7% hard / 10% soft) are still enforced by the kubelet at runtime, but the capacity annotation is not rendered. Without this annotation, the cluster-autoscaler has no visibility into these reservations and may over-schedule the node until evictions fire.

Example: Override kubelet reservations

nodeGroups:
  md0:
    instanceType: "u1.large"
    kubelet:
      systemReservedMemory: "256Mi"
      kubeReservedMemory: "256Mi"
      evictionHardMemory: "500Mi"
      evictionSoftMemory: "1Gi"

U Series: Universal

The U Series is quite neutral and provides resources for general purpose applications.

U is the abbreviation for "Universal", hinting at the universal attitude towards workloads.

VMs of instance types will share physical CPU cores on a time-slice basis with other VMs.

U Series Characteristics

Specific characteristics of this series are:

• Burstable CPU performance - The workload has a baseline compute performance but is permitted to burst beyond this baseline, if excess compute resources are available.
• vCPU-To-Memory Ratio (1:4) - A vCPU-to-Memory ratio of 1:4, for less noise per node.

O Series: Overcommitted

The O Series is based on the U Series, with the only difference being that memory is overcommitted.

O is the abbreviation for "Overcommitted".

O Series Characteristics

Specific characteristics of this series are:

• Burstable CPU performance - The workload has a baseline compute performance but is permitted to burst beyond this baseline, if excess compute resources are available.
• Overcommitted Memory - Memory is over-committed in order to achieve a higher workload density.
• vCPU-To-Memory Ratio (1:4) - A vCPU-to-Memory ratio of 1:4, for less noise per node.

CX Series: Compute Exclusive

The CX Series provides exclusive compute resources for compute intensive applications.

CX is the abbreviation of "Compute Exclusive".

The exclusive resources are given to the compute threads of the VM. In order to ensure this, some additional cores (depending on the number of disks and NICs) will be requested to offload the IO threading from cores dedicated to the workload. In addition, in this series, the NUMA topology of the used cores is provided to the VM.

CX Series Characteristics

Specific characteristics of this series are:

• Hugepages - Hugepages are used in order to improve memory performance.
• Dedicated CPU - Physical cores are exclusively assigned to every vCPU in order to provide fixed and high compute guarantees to the workload.
• Isolated emulator threads - Hypervisor emulator threads are isolated from the vCPUs in order to reduce emaulation related impact on the workload.
• vNUMA - Physical NUMA topology is reflected in the guest in order to optimize guest sided cache utilization.
• vCPU-To-Memory Ratio (1:2) - A vCPU-to-Memory ratio of 1:2.

M Series: Memory

The M Series provides resources for memory intensive applications.

M is the abbreviation of "Memory".

M Series Characteristics

Specific characteristics of this series are:

• Hugepages - Hugepages are used in order to improve memory performance.
• Burstable CPU performance - The workload has a baseline compute performance but is permitted to burst beyond this baseline, if excess compute resources are available.
• vCPU-To-Memory Ratio (1:8) - A vCPU-to-Memory ratio of 1:8, for much less noise per node.

RT Series: RealTime

The RT Series provides resources for realtime applications, like Oslat.

RT is the abbreviation for "realtime".

This series of instance types requires nodes capable of running realtime applications.

RT Series Characteristics

Specific characteristics of this series are:

• Hugepages - Hugepages are used in order to improve memory performance.
• Dedicated CPU - Physical cores are exclusively assigned to every vCPU in order to provide fixed and high compute guarantees to the workload.
• Isolated emulator threads - Hypervisor emulator threads are isolated from the vCPUs in order to reduce emaulation related impact on the workload.
• vCPU-To-Memory Ratio (1:4) - A vCPU-to-Memory ratio of 1:4 starting from the medium size.