Trilio for Kubernetes Preflight Checks Plugin

Learn how to run Trilio's Preflight checks before installation

tvk-preflight is a kubectl plugin which checks if all the prerequisites are met before installing the Trilio application in a Kubernetes cluster.

This plugin automatically generates a log file for each preflight run(preflight-<date>T<time>.log) and cleanup run (preflight_cleanup-<date>T<time>.log)

Pre-requisites:

  1. krew[optional] - kubectl-plugin manager. Install from here.

  2. kubectl - kubernetes command-line tool. Install from here.

  3. bash(>=v3.2.x) should be present on system.

For OpenShift environments, if kubectlis not installed and oc binary can be used to perform the same steps. A symbolic link can be created as shown:

sudo ln -s /usr/local/bin/oc /usr/local/bin/kubectl
  • Verify oc installation location with the command 'which oc'

  • To delete/remove symbolic links use unlink:

    unlink /usr/local/bin/kubectl

Supported OS:

  • linux/amd64

  • linux/x86

  • linux/arm

  • linux/arm64

  • darwin/amd64

  • darwin/arm64

Checks Performed during Preflight

Preflight plugin performs checks on system where this plugin is installed and few checks are performed on the kubernetes cluster where current-context of kubeconfig is pointing to.

Whenever a preflight check is performed, a 6 digit, lowercaseUID is generated for that particular preflight check. This UID is the value of the label preflight-run, which is set on every resource created during the preflight check. Also the UID is the suffix of name of every resource created during preflight check. This UID is particularly useful to perform cleanup of resources created during a particular preflight check. The following checks are included in preflight:

  1. check-kubectl - Ensures kubectl utility is present on system. This check is skipped if the --in-cluster flag is enabled.

  2. check-cluster-access - Ensures preflight can access the remote target cluster.

  3. check-helm-version

    1. Ensures helm[version>=v3.x.x] utility is present on system and pointed to the cluster.

    2. Aborts successfully for Openshift cluster.

    3. This check is skipped if --in-cluster flag is enabled.

  4. check-kubernetes-version - Ensures minimum Kubernetes version >= 1.19.x

  5. check-kubernetes-rbac - Ensures RBAC is enabled in cluster

  6. check-csi

    1. Checks if the following CSI apis are present on the cluster:

      • "volumesnapshotclasses.snapshot.storage.k8s.io"

      • "volumesnapshotcontents.snapshot.storage.k8s.io"

      • "volumesnapshots.snapshot.storage.k8s.io"

    2. If not present, the missing CSI apis is created, as per the k8s server version. If k8s server version is 1.19, the above CSI apis are installed that support v1beta1 version. If k8s server version is 1.20+, the above CSI apis installs that support both v1 and v1beta1 version. Also, if volumesnapshot CRDs don't exist, any provided volume snapshot class will be overridden with default value.

  7. check-storage-snapshot-class -

    1. Ensures that provided storageClass is present in cluster:

      1. Provided storageClass's provisioner [JSON Path: storageclass.provisioner] should match with provided volumeSnapshotClass's driver[JSON Path: volumesnapshotclass.driver]

      2. If volumeSnapshotClass is not provided then, volumeSnapshotClass which satisfies condition [i] will be selected. If there's are multiple volumeSnapshotClasses satisfying condition [i], default volumeSnapshotClass[which has annotation snapshot.storage.kubernetes.io/is-default-class: "true" set] will be used for further pre-flight checks.

      3. If volumeSnapshotClass is provided and matches with storage class provisioner, only then that volumeSnapshotClass will be used for further operations, otherwise preflight will fail with not found error.

    2. Ensures at least one volumeSnapshotClass is marked as default in cluster if user has not provided volumeSnapshotClass as input.

  8. check-dns-resolution -

    1. Ensure DNS resolution works as expected in the cluster:

      • Creates a new pod (dnsutils-${UID}) then resolves kubernetes.default service from inside the pod

  9. check_volume_snapshot -

    1. Ensure Volume Snapshot functionality works as expected for both mounted and unmounted PVCs

      1. Creates a Pod and PVC (source-pod-${UID} and source-pvc-${UID}).

      2. Creates Volume snapshot (snapshot-source-pvc-${UID}) from the mounted PVC(source-pvc-${UID}).

      3. Creates volume snapshot of unmounted PVC(source-pvc-${UID} [deletes the source pod before snapshotting].

      4. Restores PVC(restored-pvc-${UID}) from volume snapshot of mounted PVC and creates a Pod(restored-pod-${UID}) and attaches to restored PVC.

      5. Restores PVC(unmounted-restored-pvc-${UID}) from volume snapshot from unmounted PVC and creates a Pod(unmounted-restored-pod-${UID}) and attaches to restored PVC.

      6. Ensure data in restored PVCs is correct[checks for a file[/demo/data/sample-file.txt] which was present at the time of snapshotting].

    2. If check-storage-snapshot-class fails then, check_volume_snapshot check is skipped.

After all above checks are performed, cleanup of all the intermediate resources created during preflight checks' execution is done.

Installation, Upgrade, Removal of Plugins :

1. With krew:

  • Add T4K custom plugin index of krew:

    kubectl krew index add tvk-plugins https://github.com/trilioData/tvk-plugins.git
  • Installation:

    kubectl krew install tvk-plugins/tvk-preflight
  • Upgrade:

    kubectl krew upgrade tvk-preflight
  • Removal:

    kubectl krew uninstall tvk-preflight

2. Without krew:

  1. Choose a version of preflight plugin to install and check if release assets have preflight plugin's package[preflight.tar.gz]

  2. Set env variable version=v1.x.x [update with your desired version]. If version is not exported, latest tagged version will be considered.

Linux/macOS

  • Bash or ZSH shells

(
  set -ex; cd "$(mktemp -d)" &&
  OS="$(uname | tr '[:upper:]' '[:lower:]')" &&
  ARCH="$(uname -m | sed -e 's/x86_64/amd64/' -e 's/\(arm\)\(64\)\?.*/\1\2/' -e 's/aarch64$/arm64/')" &&
  if [[ -z ${version} ]]; then version=$(curl -s https://api.github.com/repos/trilioData/tvk-plugins/releases/latest | grep -oP '"tag_name": "\K(.*)(?=")'); fi &&
  echo "Installing version=${version}" &&
  package_name="preflight_${version}_${OS}_${ARCH}.tar.gz" &&
  curl -fsSLO "https://github.com/trilioData/tvk-plugins/releases/download/"${version}"/${package_name}" &&
  tar zxvf ${package_name} && sudo mv preflight /usr/local/bin/kubectl-tvk_preflight
)

You can verify the installation using the following command:

kubectl tvk-preflight --help

Windows

NOT SUPPORTED

Note for Dark Site Installation

  • For using the local-registry flag, it is mandatory to have busybox:latest and dnsutils:1.3 images (with the same tags) in the private registry.

    Steps for pushing images to local registry

    • Pull the images (dnsutils:1.3 & busybox) to local machine.

    • use the following command to push it to the local registry

    • docker push <local registry/image>

    • Example: docker push localhost:5000/busybox

Usage

kubectl tvk-preflight [sub-command] [flags]

The preflight binary has three common flags to both the subcommands.

Common Flags

--in-cluster flag should only be set when running inside a container where kubectl and helm checks are not required.\

Inputs for Preflight Checks & Cleanup

The inputs for running preflight checks and cleanup can be provided through a single file. The format of data in a file should be according to the below example:

run:
  storageClass: <storage-classs>
  snapshotClass: <snapshot-class>
  namespace: <perform preflight checks in the given namespace>
  kubeconfig: <kubeconfig file path>
  inCluster: <bool>
  serviceAccount: <service-account>
  localRegistry: <complete path of the registry to pull the images from>
  imagePullSecret: <Name of the secret while pulling images from the local registry>
  cleanupOnFailure: <Boolean. If true cleans the preflight resources after a failed preflight run>
  pvcStorageRequest: <Storage request value of PVC for volume snapshot check>
  resources:
    requests:
      memory: <pod memory request for snapshot check, e.g 64Mi>
      cpu: <pod cpu request for snapshot check, e.g 250m>
    limits:
      memory: <pod memory limit for snapshot check, e.g 128Mi>
      cpu: <pod cpu limit for snapshot check, e.g 500m>
cleanup:
  namespace: <clean preflight in a particular namespace>
  kubeconfig: <kubeconfig file path>
  logLevel: <specify logging level for cleanup>
  cleanupMode: <specify the cleanup mode as 'all' or 'uid'>
  uid: <This field is used when the cleanup mode is 'uid'>
  • The cleanupMode field can have two values-

    • all - This will clean all the preflight resources present in the given namespace.

    • uid - This will clean resources of preflight with the given uid in the given namespace.

  • The user can override the values given in file using CLI flags.

  • The input fields should be present in the correct hierarchical order. An incorrect key or input field will result in an error and preflight checks will error out.

Preflight Checks

Run a preflight check with predefined values using a sample file. Download the file using below commands:

By wget:

wget https://raw.githubusercontent.com/trilioData/tvk-plugins/main/docs/preflight/sample_file.yaml

By curl:

curl https://raw.githubusercontent.com/trilioData/tvk-plugins/main/docs/preflight/sample_file.yaml -o sample_file.yaml

Examples

With --namespace:

kubectl tvk-preflight [sub-command] [sub-command flags] --namespace <namespace of the cluster>

Using shorthand notation:

kubectl tvk-preflight [sub-command] [sub-command flags] -n <namespace of the cluster>

Preflight Binary Subcommands

There are two subcommands to the preflight binary:

  • run - To perform preflight checks.

  • cleanup - To clean the resources generated during failed preflight checks.

1. run

run sub-command performs the actual preflight checks on the system and on the Kubernetes cluster where the system's kubeconfig is pointing to in the given namespace. The following table lists the associated flags:

Examples:

Storage class is a required flag for the run subcommand. With

With --volume-snapshot-class: Performs preflight checks on the cluster with the given volumeSnapshotClass in the given namespace.

kubectl tvk-preflight run --storage-class <storageclass name> --volume-snapshot-class <volumeSnapshotClass name>

Pod Scheduling

The pods of preflight run can be made to schedule on a particular set of nodes of cluster by specifying the labels for node selection, node affinity, pod affinity/anti-affinity and taints and toleration.

The labels except node-selector can only be specified through a config file of preflight run. Currently, it is not possible to specify pod and node affinity and tolerations through CLI flags.

Please refer below examples for specifying labels onto the pods of preflight run.

Examples

  run:
  ...
  podSchedulingOptions:
    nodeSelector:
      node-sel-key: node-sel-value
  ...

2. cleanup

  • cleanup subcommand cleans/deletes the resources created during failed preflight checks and not cleaned up on failure.

  • The cleanup command will clean all the resources generated due to preflight checks in the given namespace.

  • Users can clean resources of a particular preflight check by specifying the uid of the preflight check.

Flags:

Examples:

Cleans all the preflight resources present on the cluster in the given namespace.

kubectl tvk-preflight cleanup --namespace <namespace of the cluster>

If namespace is not specified then, cleanup will be performed in default namespace of the cluster.