> ## Documentation Index
> Fetch the complete documentation index at: https://formbricks.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# 5.4: Dependency changes

> Valkey, Cube and Envoy actions for the 5.4 release line.

**Applies to:** dependency changes across the 5.4 release line. Follow the version boundaries below.

[Back to the 5.4 checklist](https://formbricks.com/docs/self-hosting/upgrade/versions/5.4.md).

## Cube from 5.4.5

The bundled image changes from upstream Cube to the Formbricks image, starting with `1.7.47-1` in 5.4.5.
Use the image shipped by your selected patch. Follow the [Cube service update](https://formbricks.com/docs/self-hosting/upgrade/versions/cube.md)
with that **5.4** tag, not a v6 Compose file.

### Bundled Valkey 8.1.9 upgrade (action required for some self-hosted deployments)

**Introduced in:** Formbricks 5.4.0, for deployments using bundled Valkey in Compose or Helm.

The old bundled digest selects only `linux/amd64`. The new digest selects the matching native image on
`linux/amd64` and `linux/arm64`, without requiring QEMU or `binfmt`. Valkey 8.1.9 also contains upstream
security fixes that are not present in 8.1.1.

> [!WARNING]
>
> Valkey stores sessions, queues, cache entries and rate-limit state in the named `redis` volume with AOF
> enabled. Do not run `docker compose down -v`, delete the volume, or use `FLUSHDB` during this upgrade.
> Recreating only the container preserves the volume. Expect a brief interruption while Valkey restarts.

Use the `redis` image selected in the target release's Compose/chart, preserving its volume/PVC and
configuration. Apply it during the maintenance procedure; validate the rendered configuration before rollout.
For Helm, update any old `redis.image.digest` override as well as the repository/tag so it does not mask the new image.

After rollout, run these from your Compose directory (`redis` is the service name):

```bash theme={null}
docker compose exec -T redis valkey-cli ping
docker compose exec -T redis valkey-server --version
docker compose ps redis
```

Expect `PONG`, the target Valkey version and a healthy container. For Kubernetes, run the same commands
inside the cache pod with `kubectl exec -n <namespace> <redis-pod> -- <command>`. Check the resolved image
supports your nodes’ architecture.
The one-click updater can replace the exact old pin, but use the release-matched updater for your target;
do not download today's `stable` script for a historical upgrade.

The chart also stops publishing the bundled cache to the host. If a component outside the deployment
previously used that port, provide a private authenticated route or external Redis instead of reopening it publicly.

### Envoy Gateway 1.8.4 CRD upgrade

**From 5.4.3**, the Helm chart replaces Envoy Gateway 1.7.1 and Gateway API 1.4.1 with Envoy Gateway 1.8.4 and Gateway
API 1.5.1. Fresh installations receive the new CRDs automatically. Helm does not update CRDs during an ordinary
upgrade, so existing installations using the bundled controller must apply the matching CRDs before upgrading:

```bash theme={null}
crd_dir="$(mktemp -d)"
helm pull oci://docker.io/envoyproxy/gateway-crds-helm \
  --version v1.8.4 --untar --untardir "${crd_dir}"
helm template envoy-gateway-crds "${crd_dir}/gateway-crds-helm" \
  --set crds.gatewayAPI.enabled=true \
  --set crds.gatewayAPI.channel=experimental \
  --set crds.envoyGateway.enabled=true \
  > "${crd_dir}/crds.yaml"
# Review the rendered CRDs and ownership before applying:
kubectl apply --server-side -f "${crd_dir}/crds.yaml"
```

If field ownership conflicts, coordinate with the existing platform owner; do not automatically force ownership.
Pulling the OCI chart first keeps Helm 4's pull messages out of the manifest pipe. Remove the temporary directory
after the apply succeeds.

The chart does not install Envoy Gateway's cluster-wide safe-upgrade admission policy by default. Enabling
`envoy.crds.gatewayAPI.safeUpgradePolicy.enabled` denies Gateway API CRD versions older than v1.5.0 across the
whole cluster, including CRDs reconciled by unrelated releases. Enable it only after platform review; a chart
rollback does not remove that policy automatically.

If the platform already manages compatible Gateway API CRDs, do not install another Gateway API bundle. Install
only the Envoy Gateway CRDs as described in the [Kubernetes guide](https://formbricks.com/docs/self-hosting/setup/kubernetes.md), set
`envoy.crds.enabled=false`, and configure Flux to skip CRD installation and upgrades. Upgrade the CRDs before the
controller and never downgrade them during rollback.

The controller scheduling values also move from the ineffective `envoy.deployment.envoyGateway.pod` path to
`envoy.deployment.pod`. Move any node selectors, tolerations, topology spread constraints, extra volumes, or
pod annotations that you placed under the old path before upgrading.

The parent chart no longer repeats the bundled controller's default port list. The
`envoy.deployment.ports` override remains supported, so existing port overrides need no migration.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.