Skip to content

Latest commit

 

History

History
137 lines (109 loc) · 4.61 KB

README.md

File metadata and controls

137 lines (109 loc) · 4.61 KB
Raw

Migrating workloads from one UKS cluster to another using Velero

This is an example of how to migrate workloads from an existing UKS cluster to a new UKS cluster.

Prerequisites

  • Your existing UKS cluster
  • A new, fresh UKS cluster (the one you want to migrate to)
  • Velero CLI
  • s3cmd

Some notes before migrating

  • Velero doesn't support restoring into a cluster with a lower Kubernetes version than where the backup was taken.
  • This guide will not cover migrating persistent volumes.
  • Going forth, the initial cluster will be referred to as "Cluster A" and the new cluster, "Cluster B".

Set up object storage bucket

We will configure Velero to use Object Storage 2.0 to store backup data from Cluster A. We will then use this to restore into Cluster B.

  • In the Hub, head to the Object Storages page and hit 'Create Object Storage'. Choose your desired region, storage size and name and hit create.

  • Then, head to the Users tab under your new Object Storage instance and click '+ Access Key'. Give it a name, and copy the resulting Access and Secret keys into a text file somewhere. We will need this to create our bucket next.

  • To create a bucket in your new Object Storage instance, we will use s3cmd. Let's use the interactive configuration wizard:

    s3cmd --configure

Access Key: {access_key}
Secret Key: {secret_key}
Default Region [US]:    # leave as default

S3 Endpoint [s3.amazonaws.com]: {hostname}

DNS-style bucket+hostname:port template for accessing a bucket [%(bucket)s.s3.amazonaws.com]: %(bucket).{hostname}

Encryption password:   # hit enter
Path to GPG program:   # leave as default

Use HTTPS protocol [Yes]:  # hit enter

HTTP Proxy server name:   # leave empty

...

Test access with supplied credentials? [Y/n]
Please wait, attempting to list all buckets...
Success. Your access key and secret key worked fine :-)

Now verifying that encryption works...
Not configured. Never mind.

Save settings? [y/N] y

  • Replace {access_key} and {secret_key} with the values copied from earlier. Replace {hostname} with the displayed hostname under the section labeled 'Public access' in your Object Storage instance. This will create an ~/.s3cfg file (in your user's home directory).

  • Finally, run

    s3cmd -c ~/.s3cfg mb s3://velero

    to create a bucket called velero. Head to the 'Buckets' tab under your Object Storage instance to verify it got created.

Performing the migration

  • First, create a config file for Velero containing your bucket's access and secret keys like so (velero.conf):

    [default]
aws_access_key_id={access_key}
aws_secret_access_key={secret_key}

  • We will use Velero's AWS S3 provider to interface with UpCloud Object Storage. Now, let's install Velero on both clusters A and B. Ensure your kubeconfigs are pointing to the correct clusters:

    export KUBECONFIG=path/to/cluster-A-kubeconfig
velero install --provider aws \
    --plugins velero/velero-plugin-for-aws:v1.8.0 \
    --bucket velero --secret-file ./velero.conf \
    --backup-location-config region=europe-1,s3ForcePathStyle="true",s3Url=https://{hostname} \
    --use-volume-snapshots=false

export KUBECONFIG=path/to/cluster-B-kubeconfig
# repeat for cluster B

  • In Cluster A, run:

    velero backup create backup-1 --include-resources='*' --wait
velero backup logs backup-1    # to inspect backup logs

  • Once done, in Cluster B, check to see if the backup we created in Cluster A is visible:

    velero backup describe backup-1

    You may need to wait a bit for it to show up since Velero in Cluster B might need to synchronize against the configured object storage bucket.

  • Restore from the backup using:

    velero restore create --from-backup backup-1
velero restore get
velero restore describe <restore-name-from-get-command>

Caveats

If Cluster A has services of type LoadBalancer, they will not come up correctly in Cluster B because the new set of service accounts cannot access the actual UpCloud Load Balancer via the API. The workaround is to recreate these services manually:

  • Delete the initially restored service. You may have to kubectl edit to remove the finalizer for the delete to go through.
  • Recreate the service with a different name than the original: 
    kubectl expose ... --type='LoadBalancer' --name='my-new-loadbalancer-service'