Skip to main content
Version: 9.0.11 (latest)

Air-Gapped Install

For environments that are connected to the Internet, one needs access to three repositories to install Veeam Kasten:

  • The Helm repository that contains the Veeam Kasten chart
  • The container registry that contains the Veeam Kasten container images
  • Upstream repositories to install Veeam Kasten dependencies (e.g., Prometheus)

However, if an air-gapped installation is required, it is possible to use your own private container registry to install Veeam Kasten. While this can always be done manually, the k10tools image command makes it easier to automate the process.

Note

An air-gapped installation has no outbound connectivity. If the cluster can reach the internet through an HTTP(S) proxy instead, install Veeam Kasten normally and configure the proxy as described in Configuring an HTTP(S) Proxy.

Air-Gapped Veeam Kasten Installation​

If the Veeam Kasten container images are already available in a private repository, the below instructions can be used to install in an air-gapped environment. If needed, support for uploading images to a private image registry is documented below.

Fetching the Helm Chart for Local Use​

To fetch the most recent Veeam Kasten Helm chart for local use, run the following command to pull the latest Veeam Kasten chart as a compressed tarball (.tgz) file into the working directory.

$ helm repo update && \
helm fetch kasten/k10

If you need to fetch a specific version, please run the following command:

$ helm repo update && \
helm fetch kasten/k10 --version=<k10-version>

Installing Veeam Kasten with Local Helm Chart and Container Images​

If the Veeam Kasten container images were uploaded to a registry at repo.example.com, an air-gapped installation can be performed by setting global.airgapped.repository=repo.example.com as shown in the below command:

$ kubectl create namespace kasten-io
$ helm install k10 k10-9.0.11.tgz --namespace kasten-io \
--set global.airgapped.repository=repo.example.com

Installing Veeam Kasten with Disconnected OpenShift Operator​

To install Veeam Kasten with an OpenShift operator in an air-gapped cluster, follow the steps under offline operator install.

Running Veeam Kasten Within a Local Network​

To run Veeam Kasten in a network without the ability to connect to the internet, Veeam Kasten needs to be installed in an air-gapped mode with the helm value metering.mode=airgap as shown in the command below:

$ kubectl create namespace kasten-io
$ helm install k10 k10-9.0.11.tgz --namespace kasten-io \
--set metering.mode=airgap
Note

If metering.mode=airgap is not set in an offline cluster, some functionality will be disabled. A message warning that Veeam Kasten is "Unable to validate license" will be displayed in the web based user interface. Errors containing messages "Could not get google bucket for metrics", "License check failed" and "Unable to validate license" will be logged.

If the metering service is unable to connect to the internet for 24 hours, the metering service will restart.

Providing Credentials if Local Container Repository is Private​

If the local repository that has been provided as the value of global.airgapped.repository is private, credentials for that repository can be provided using secrets.dockerConfig and global.imagePullSecret flags, as below, with the helm install command.

--set secrets.dockerConfig=$(base64 -w 0 < ${HOME}/.docker/config.json) \
--set global.imagePullSecret="k10-ecr"
Note

Our Helm chart creates a secret with the name k10-ecr with the value that has been provided for secrets.dockerConfig. That's why we are providing secret name k10-ecr as value of global.imagePullSecret.

Enabling Kanister Operations on Kubernetes 1.35 or Later​

On Kubernetes version 1.35 or later, a pod must present its own registry credentials to use a container image, even one already present on the node. Some Veeam Kasten worker pods run in the application namespace, where the k10-ecr secret does not exist, and so cannot pull images from a private repository.

When you use a private repository, this affects the following operations:

  • Backup and restore of applications protected by a Kanister blueprint
  • Backup of volumes that are not attached to a running workload
  • Execution hooks whose blueprint creates a pod

The affected operations can be enabled in following two ways.

Manually copy image pull secrets to application namespaces​

To enable these operations, copy the k10-ecr secret into each protected namespace and attach it to the default service account of that namespace. For more about image pull secrets, see Pull an Image from a Private Registry.

$ export APP_NAMESPACE=<application-namespace>

$ kubectl get secret k10-ecr --namespace=kasten-io --output=json \
| jq '{apiVersion, kind, metadata: {name: .metadata.name}, type, data}' \
| kubectl apply --namespace=${APP_NAMESPACE} --filename=-

$ kubectl patch serviceaccount default \
--namespace=${APP_NAMESPACE} \
--patch='{"imagePullSecrets":[{"name":"k10-ecr"}]}'
Note
  • If you provided a name other than k10-ecr for global.imagePullSecret, use that name instead.
  • The patch replaces the imagePullSecrets field of the service account instead of merging into it. If the default service account already lists image pull secrets, include them in the patch as well.
  • To restore an application into a namespace that does not exist yet, create and configure that namespace before running the restore.

Configure Veeam Kasten to create image pull secret copies​

Alternatively, Veeam Kasten can be configured to create the copies of image pull secrets in application namespaces for the duration of operation. To enable this, set flag global.copyImagePullSecretsToAppNamespaces to true with helm install command as follows.

--set global.copyImagePullSecretsToAppNamespaces=true

When configured, Veeam Kasten creates copies of secrets specified using global.imagePullSecrets in application namespaces. The secrets are automatically deleted when operation finishes.

The automatic creation of temporary image pull secrets is currently supported for following scenarios:

  • Backup and restore of applications protected by a Kanister blueprint
  • Backup of volumes that are not attached to a running workload

Enabling Veeam Kasten Primer Operations on Kubernetes 1.35 or later​

Veeam Kasten Primer operations are also affected by issue described in Enabling Kanister Operations on Kubernetes 1.35 or Later. Following sections describe steps to enable various primer operations.

Enabling operations by primer script and k10tools primer gvs-cluster-check subcommand​

Veeam Kasten Primer script k10_primer.sh offers pre-flight checks as described in Installation Pre-flight Checks, CSI Pre-flight Checks, and Veeam Kasten Primer Block Mount Check. To let Veeam Kasten to be able to perform these operations correctly, k10-ecr secret must be present in default namespace and default service account must be patched to use the k10-ecr secret. If an operation allows passing a custom namespace and service account, make sure namespace being passed contains k10-ecr secret and the service account must be configured to use the k10-ecr. Steps mentioned in Manually copy image pull secrets to application namespaces can help setting this up. Make sure values for namespace and service account are adjusted accordingly.

warning

Currently, only following k10tools primer subcommands will work by adding secrets to namespace and patching service account.

  • k10tools primer storage check csi
  • k10tools primer storage check blockmount
  • k10tools primer gvs-cluster-check

Cloud provider integration check subcommands k10tools primer storage check <awsebs|azure|gcepd> do not work with method described above.

Enabling storageclass validation from Veeam Kasten UI​

Veeam Kasten UI provides way to validate storageclass from Settings -> System Information page. In air-gapped environments on Kubernetes v1.35 and above, the default service account of the namespace where Veeam Kasten is installed (default kasten-io), needs to be configured to use k10-ecr secret. Refer to instructions in Manually copy image pull secrets to application namespaces to configure service account to use secret. Make sure to adjust namespace name to kasten-io (or the namespace where Veeam Kasten is installed) while executing the command.

Preparing Veeam Kasten Container Images for Air-Gapped Use​

There are multiple ways to use a private repository including setting up a caching or proxy image registry that points to the Veeam Kasten image repositories using tools such as JFrog Artifactory. However, if images need to be manually uploaded or an automated upload pipeline is required to add Veeam Kasten images into your private repository, the following documentation should help.

To see all available commands and flags for running k10tools image please run the following:

$ docker run --rm gcr.io/kasten-images/k10tools:9.0.11 image --help

The following commands operate against the latest version of Veeam Kasten (9.0.11).

warning

k10tools image is only supported for versions 7.5.0+ of Veeam Kasten and must match the version you're installing.

For older version, please refer to their documentation: https://docs.kasten.io/<version>/install/offline.html.

List Veeam Kasten Container Images​

The following command will list all images used by the current Veeam Kasten version (9.0.11). This can be helpful if there is a requirement to tag and push Veeam Kasten images into your private repository manually instead of using the Kasten provided tool documented below.

$ docker run --rm gcr.io/kasten-images/k10tools:9.0.11 image list

Copy Kasten Images into a Private Repository​

The following command will copy the Veeam Kasten container images into your specified registry. If the destination image tag should be different than the Veeam Kasten version, then the --dst-image-tag can be used to specify a new image tag.

The following example uses a repository located at repo.example.com.

$ docker run --rm -v $HOME/.docker:/home/kio/.docker gcr.io/kasten-images/k10tools:9.0.11 image copy --dst-registry repo.example.com
Note

This command will use your local docker config if the private registry requires authentication.

The credsStore field in the $HOME/.docker/config.json is used to specify the credential store. This is typically an external credential store requiring an external helper and it may not be usable from within the docker container. Please refer to the docker documentation for more information.

Alternatively, k10tools image provides authentication mechanisms such as passing a username and password (--dst-username and --dst-password flags) or a bearer token (--dst-token flag). Please refer to the help flag for more information.

After running the previous command, use the instructions above to install Veeam Kasten via images uploaded to repo.example.com.

Copy Kasten Images to/from a Filesystem Directory​

Network limitations may limit the ability to directly copy images into a private repository. Alternatively, images can be copied to the local filesystem and then pushed to a repository separately. This requires downloading the k10tools binary.

The following example copies the images to a directory images. This directory can then be used to upload to a private repository located at repo.example.com.

:substitutions:
$ k10tools image copy --dst-path images
$ k10tools image copy --src-path images --dst-registry repo.example.com

Using Iron Bank Veeam Kasten Container Images​

If you want to use the Iron Bank hardened Veeam Kasten images in an air-gapped environment, execute the above commands but replace image with ironbank image:

:substitutions:
$ docker run --rm gcr.io/kasten-images/k10tools:9.0.11 ironbank image list
$ docker run --rm -v $HOME/.docker:/home/kio/.docker gcr.io/kasten-images/k10tools:9.0.11 ironbank image copy --dst-registry repo.example.com

This ensures the images are pulled from Registry1.

warning

You must be logged in to the docker registry locally for this process to function correctly. Use docker login registry1.dso.mil --username "${REGISTRY1_USERNAME}" --password-stdin with your Registry1 CLI secret as the password to login.

Alternatively, provide credentials using the methods described above.