Updating

This guide will show you how to update your Coder deployment.

Prerequisites

• If you haven't already, install helm.

• Before beginning the update process, ensure that you've added the Coder helm repo to your cluster. You can verify that the Coder repo has been added to helm using helm repo list:

  $ helm repo list
  NAME URL
  coder https://helm.coder.com
  

  If you don't have the Coder repo, you can add it:

  helm repo add coder https://helm.coder.com
  

• Ensure that you have superuser privileges to your PostgreSQL database.

Recommendations

• As with any significant maintenance operation, we strongly recommend taking a snapshot of the database before proceeding with the upgrade. In the event that there are upgrade issues, it is simpler and safer to roll back directly at the database level, since it guarantees restoration of the system to a known-working condition.

• We recommend updating no more than one major version at a time (i.e., we recommend moving from 1.15 to 1.16 only).

Update Coder

To update Coder, follow these steps:

1. Retrieve the latest repository information:

   helm repo update
   

2. (Optional) Export the current helm chart values into a file:

   helm get values --namespace coder coder > current-values.yml
   

   Make sure that your values only contain the changes you want (i.e., if you see references to a prior version, you may need to remove these).

3. Provide your helm chart values file and upgrade to the desired version (e.g., 1.16.1):

   Note: If you omit --version, you'll upgrade to the latest version.

   helm upgrade --namespace coder --install --atomic --wait \
     --version 1.16.1 coder coder/coder --values current-values.yml
   

Fixing a Failed Upgrade

While upgrading, the process may fail. You'll see an error message similar to the following samples indicating that a field is immutable or that helm doesn't control a resource:

failed to replace object: Service "cemanager" is invalid:
spec.clusterIP: Invalid value: "": field is immutable

Error: UPGRADE FAILED: rendered manifests contain a resource
that already exists. Unable to continue with update:
ServiceAccount "coder" in namespace "coder" exists and cannot
be imported into the current release: invalid ownership metadata;
label validation error: missing key
"app.kubernetes.io/managed-by": must be set to "Helm"; annotation
validation error: missing key "meta.helm.sh/release-name": must
be set to "coder"; annotation validation error: missing key
"meta.helm.sh/release-namespace": must be set to "coder"

If this happens, we recommend uninstalling and reinstalling:

1. Export the helm chart values into a file:

   helm get values --namespace coder coder > current-values.yml
   

   Double-check your values file to ensure it only contains your changes.

2. Run helm uninstall:

   helm uninstall --namespace coder coder
   

   Make sure to check the namespace for items that are slow to delete. For example web-ingress can take some time to release the IP addresses; if you run the install command before this process completes, the install process will fail.

   Running uninstall removes the web-ingress service that owns the ingress controller's IP address. As such, the web-ingress service may have a new public IP address or hostname after you reinstall; if this is the case, update your DNS provider with your new IP and CNAME.

3. Run the upgrade command with the new version number and helm chart values file:

   helm upgrade --namespace coder --atomic \
   --wait --install --version 1.16.1 \
   coder coder/coder --values current-values.yml
   

   The ingress may attach to a new public IP address; if this happens, you must update the host and Dev URL IP addresses with your DNS provider.