# Upgrade Steps

- Updated on Feb 20, 2026
- Published on Sep 17, 2024

- 4 minute(s) read

## Prerequisites

Before upgrading the application, you need to stream the new version to your Docker Registry. If you need to install `hsk8s`, refer to the [hsk8s (Hyperscience Kubernetes CLI)](https://help.hyperscience.ai/deployment/docs/hsk8s-hyperscience-kubernetes-cli) article.

```bash
export HS_APP_VERSION=<version>
export HS_TOKEN=<token>

hsk8s image stream your-docker-registry.net --token $HS_TOKEN --aws
hsk8s image stream your-docker-registry.net --token $HS_TOKEN --aws --forms --forms-repo  --forms-version $HS_APP_VERSION
hsk8s image stream your-docker-registry.net --token $HS_TOKEN --aws --blocks --blocks-repo   --forms-version $HS_APP_VERSION
hsk8s image stream your-docker-registry.net --token $HS_TOKEN --aws --trainer --trainer-repo  --forms-version $HS_APP_VERSION
```

**Note** Replace `<version>`and `<token>` with actual values. You should have received your token from Hyperscience sales representative.

**Note** Replace `your-docker-registry.net` with your Docker registry endpoint.

Make sure that you have imported environment variables from [hsk8s (Hyperscience Kubernetes CLI)](https://help.hyperscience.ai/deployment/docs/hsk8s-hyperscience-kubernetes-cli#installing-kubectl):

```bash
source hs_env.bash
```

## Minor Upgrade

Update the image tags in your `values.yaml` file with the new forms version that you want to install.

```yaml
app:
  tag: ""

trainer:
  -
```

**Note** You need kubectl access to run the following command. Once you have saved your changes run the following helm command to upgrade the release:

`helm upgrade $HS_HELM_RELEASE -f values.yaml $HS_HELM_CHART`

This will kick-off the upgrade process, during which you can expect some downtime.

## Major Upgrade

> **IMPORTANT:** When performing a major version upgrade on the Hyperscience application, a new version of the trainer needs to be spawned first to maintain automation performance. More information on that topic can be found in the [How Upgrades Work in Hypercell](https://help.hyperscience.ai/deployment/docs/how-upgrades-work-in-hypercell#training-models).

```yaml
trainer:
  tags:
  -
  -
```

Apply the changes `$ helm upgrade $HS_HELM_RELEASE -f values.yaml $HS_HELM_CHART`

Two new trainer pods will be started and once the artifact jobs are finished the following changes are to be made to the `values.yaml` file

```yaml
app:
  tag: ""

trainer:
  -  # Delete the  item
```

Apply the changes `$ helm upgrade $HS_HELM_RELEASE -f values.yaml $HS_HELM_CHART`

## Step by step sample

Initially a new init job will be created. Both the frontend and backend pods will be stuck in `Init` until the `init-job` is completed.

```yaml
$ helm upgrade $HS_HELM_RELEASE -f values.yaml $HS_HELM_CHART
$ kubectl get pod
NAME                                          READY   STATUS              RESTARTS   AGE
hs-block-lambdapython-77677cd4f6-v5cnk        1/1     Running             0          18h
hyperscience-backend-7c7777d9dd-l4nb5         0/7     Init:0/1            0          49s
hyperscience-frontend-6984bb49c4-g557k        0/2     Init:0/1            0          46s
hyperscience-hyperoperator-68f8cfcb57-79wvs   1/1     Running             0          46s
hyperscience-init-job-kaxsg-5nhwk             0/3     Init:0/2            0          41s
trainer-b683d4f1-bmbdq                        0/9     ContainerCreating   0          41s
```

**Note** Some restarts of the trainer pod are expected until the init job completes.

```yaml
$ kubectl get pod
NAME                                          READY   STATUS             RESTARTS   AGE
hs-block-lambdapython-77677cd4f6-v5cnk        1/1     Running            0          18h
hyperscience-backend-7c7777d9dd-l4nb5         0/7     Init:0/1           0          3m57s
hyperscience-frontend-6984bb49c4-g557k        0/2     Init:0/1           0          3m54s
hyperscience-hyperoperator-68f8cfcb57-79wvs   1/1     Running            0          3m54s
hyperscience-init-job-kaxsg-5nhwk             3/3     Running            0          3m49s
trainer-b683d4f1-bmbdq                        3/9     CrashLoopBackOff   8          3m49s
```

Once the init job completes and the backend pod is started hyperoperator will update all needed blocks.

```yaml
$ kubectl get pod
NAME                                          READY   STATUS             RESTARTS   AGE
hs-block-lambdapython-55c5b997fb-kbz8s        1/1     Running            0          39s
hs-block-lambdapython-77677cd4f6-v5cnk        1/1     Terminating        0          18h
hyperscience-backend-7c7777d9dd-l4nb5         7/7     Running            0          5m22s
hyperscience-frontend-6984bb49c4-g557k        2/2     Running            0          5m19s
hyperscience-hyperoperator-68f8cfcb57-79wvs   1/1     Running            0          5m19s
hyperscience-init-job-kaxsg-5nhwk             0/3     Completed          0          5m14s
trainer-b683d4f1-bmbdq                        8/9     Running   22         5m14s
```