LXD CSI driver reference

This document contains reference information for the LXD CSI driver CLI, including available Helm chart values and its versioning information.

CLI

The lxd-csi-driver provides CSI controller and node server functionality. You can configure runtime options using these flags:

Flag	Default	Description
--driver-name	lxd.csi.canonical.com	CSI driver name
--endpoint	unix:///tmp/csi.sock	Internal CSI Unix socket path
--devLXDEndpoint	unix:///dev/lxd/sock	DevLXD Unix socket path
--nodeID	""	Kubernetes node ID

Helm chart

The LXD CSI driver Helm chart is available as an OCI image. The source of the Helm chart can be found in the LXD CSI driver repository.

The table below contains configurable Helm chart values with their default values and descriptions.

Key	Type	Default	Description
driver.image.repository	string	ghcr.io/canonical/lxd-csi-driver	LXD CSI image
driver.image.tag	string	Chart version	LXD CSI image tag
driver.image.pullPolicy	string	IfNotPresent	LXD CSI image pull policy
driver.imagePullSecrets	list	[]	LXD CSI image pull secrets
driver.tokenSecretName	string	lxd-csi-secret	Name of the secret containing DevLXD bearer token in token field
driver.fsGroupPolicy	string	File	Controls Kubernetes fsGroup behavior for the driver (None, ReadWriteOnceWithFSType, File)
driver.volumeNamePrefix	string	csi	Prefix used for LXD volume names. Resulting volume name is in format <prefix>-<uuid>.
rbac.create	bool	true	Create RBAC resources allowing LXD CSI access to relevant Kubernetes objects
controller.name	string	lxd-csi-controller	Controller Deployment name
controller.replicas	int	1	Controller Deployment replicas. When deployed in multiple replicas, the preferred affinity is configured in an attempt to distribute Pods across multiple nodes.
controller.strategy.type	string	RollingUpdate	Controller Deployment update strategy (RollingUpdate, Recreate)
controller.strategy.rollingUpdate.maxUnavailable	string	50%	Max unavailable Pods during update
controller.priorityClassName	string	system-cluster-critical	Controller Pod scheduling priority
controller.serviceAccount.create	bool	true	Whether to create required service account for controller server
controller.serviceAccount.name	string	"" (Equal to Controller server name if empty)	Custom service account name
controller.runOnControlPlaneOnly	bool	true	Whether to run controller only on control plane nodes. This ensures appropriate node affinity and tolerations are configured. If node affinity is manually set, this option is disabled.
controller.nodeSelector	object	{}	Node selector for controller Pods
controller.tolerations	list	[]	Controller Pod tolerations
controller.affinity	object	{}	Controller Pod affinity
controller.annotations	object	{}	Controller Deployment annotations
controller.podAnnotations	object	{}	Controller Pod annotations
controller.resources	object	{}	Controller resource limits and requests
controller.csiProvisioner.image.repository	string	registry.k8s.io/sig-storage/csi-provisioner	CSI provisioner image
controller.csiProvisioner.image.tag	string	Chart release dependent	CSI provisioner image tag
controller.csiProvisioner.image.pullPolicy	string	IfNotPresent	CSI provisioner pull policy
controller.csiProvisioner.resources	object	{}	CSI provisioner resource limits and requests
controller.csiAttacher.image.repository	string	registry.k8s.io/sig-storage/csi-attacher	CSI attacher image
controller.csiAttacher.image.tag	string	Chart release dependent	CSI attacher image tag
controller.csiAttacher.image.pullPolicy	string	IfNotPresent	CSI attacher image pull policy
controller.csiAttacher.resources	object	{}	CSI attacher resource limits and requests
controller.csiLivenessProbe.image.repository	string	registry.k8s.io/sig-storage/livenessprobe	CSI liveness probe image
controller.csiLivenessProbe.image.tag	string	Chart release dependent	CSI liveness probe image tag
controller.csiLivenessProbe.image.pullPolicy	string	IfNotPresent	CSI liveness probe image pull policy
controller.csiLivenessProbe.resources	object	{}	CSI liveness probe resource limits and requests
node.name	string	lxd-csi-node	Node DaemonSet name
node.strategy.type	string	RollingUpdate	Node DaemonSet update strategy (RollingUpdate, OnDelete)
node.strategy.rollingUpdate.maxUnavailable	string	1	Max unavailable Pods during update
node.priorityClassName	string	system-node-critical	Node Pod scheduling priority
node.serviceAccount.create	bool	true	Whether to create required service account for node server
node.serviceAccount.name	string	"" (Equal to Node server name if empty)	Custom service account name
node.nodeSelector	object	{}	Node selector for node server Pods
node.tolerations	list	[]	Node Pod tolerations
node.affinity	object	{}	Node Pod affinity
node.annotations	object	{}	Node DaemonSet annotations
node.podAnnotations	object	{}	Node Pod annotations
node.resources	object	{}	Node server resource limits and requests
node.nodeDriverRegistrar.image.repository	string	registry.k8s.io/sig-storage/csi-node-driver-registrar	Node driver registrar image
node.nodeDriverRegistrar.image.tag	string	Chart release dependent	Node driver registrar image tag
node.nodeDriverRegistrar.image.pullPolicy	string	IfNotPresent	Node driver registrar image pull policy
node.nodeDriverRegistrar.resources	object	{}	Node driver registrar resource limits and requests
node.csiLivenessProbe.image.repository	string	registry.k8s.io/sig-storage/livenessprobe	CSI liveness probe image
node.csiLivenessProbe.image.tag	string	Chart release dependent	CSI liveness probe image tag
node.csiLivenessProbe.image.pullPolicy	string	IfNotPresent	CSI liveness probe image pull policy
node.csiLivenessProbe.resources	object	{}	CSI liveness probe resource limits and requests
storageClasses[].create	bool	true	Create the specified storage class
storageClasses[].name	string	""	Storage class name
storageClasses[].storagePool	string	""	Name of the target LXD storage pool
storageClasses[].volumeBindingMode	string	WaitForFirstConsumer	Volume binding mode (Immediate, WaitForFirstConsumer)
storageClasses[].reclaimPolicy	string	Delete	Volume reclaim policy (Delete, Retain)
storageClasses[].annotations	object	{}	Additional storage class annotations

Versioning

The LXD CSI driver follows Semantic Versioning independently of LXD releases, starting at 0.0.1.

Tag type	Format	Example	Description
Patch	v<major>.<minor>.<patch>	v1.2.3	Bugfix release. Each patch within the same minor version provides the same set of features.
Minor	v<major>.<minor>	v4.5	Feature release. Adds new features in a backward-compatible way within the major version. Features may be deprecated, but remain usable.
Major	v<major>	v6	Breaking release. May include breaking changes and may raise minimum LXD version.

Compatibility

For any major driver version ≥1, the minimum supported LXD version is fixed. Updates within that major version remain compatible with that LXD version (or newer) until the driver version reaches end-of-life (EOL).

On the other hand, Kubernetes versions receive roughly one year of patch support. The LXD CSI driver only supports Kubernetes versions that are themselves supported upstream. When a Kubernetes version reaches end of life, it is no longer supported by the driver.

CSI Version	Min. LXD Version	Min. Kubernetes Version	EOL
1	6.6	v1.31	N/A

Special versions

• All versions before the first major release (<1.0.0) make no guarantees. Behavior may change even in patch releases.

• Versions with non-semantic tags (e.g. latest-edge) or with pre-release identifiers (e.g. v1.2.3-edge) are considered unstable and should be used only for testing.

• Version 0.0.0 is reserved for unstable builds where semantic versioning is required (e.g. v0.0.0-latest-edge is latest Helm chart release).

Related topics

Explanation:

• The LXD CSI driver

How-to guides:

• How to use the LXD CSI driver with Kubernetes