---
title: Migration and Disaster Recovery (Kubernetes)
url: https://www.tines.com/stories/docs/self-hosted/best-practices/migration-and-disaster-recovery-kubernetes/
updated: 2026-10-26T23:40:18+00:00
---

*[tines.com](https://www.tines.com/llms.txt) › [Docs](https://www.tines.com/stories/docs/llms.txt) › [Self-Hosted](https://www.tines.com/llm/stories/docs/self-hosted.md) › [Best Practices](https://www.tines.com/llm/stories/docs/self-hosted/best-practices.md)*

# Migration and Disaster Recovery (Kubernetes)

*[View on tines.com](https://www.tines.com/stories/docs/self-hosted/best-practices/migration-and-disaster-recovery-kubernetes/)*

Use this to bring your Tines installation back after a cluster is lost, or to move it to a new database. Restore your database, install Tines against it, and empty your cache before Tines starts.

The backup and restore commands are in [Data backup recommendations](https://www.tines.com/stories/docs/self-hosted/best-practices/data-backup-recommendations/). Installing the chart is covered in [Helm Charts](https://www.tines.com/stories/docs/self-hosted/getting-started/helm-charts/).

> **TIP:** We highly recommend that you run PostgreSQL on a managed service, such as Amazon Aurora PostgreSQL or Azure Database for PostgreSQL. You restore that service from its own backup, and your cluster can be replaced without moving database files.

### Restore your database

Scale your running installation to zero first, so nothing writes during the restore. [Infrastructure maintenance](https://www.tines.com/stories/docs/self-hosted/upgrade-tines/infrastructure-maintenance/) uses these deployment names. Check `kubectl get deploy -n tines` and use the names you see there.

```bash
kubectl scale deploy -n tines --replicas=0 tines-app tines-sidekiq
```

Restore your latest backup to a new PostgreSQL instance and wait until it accepts connections. A copied data directory has to go back onto the same PostgreSQL major version. A `pg_dump` file can go onto that version or a newer one.

Deploy the same Tines version you were running, or a newer one. On startup, Tines migrates an older database. A newer database will not run on an older Tines.

### Install Tines against your restored database

Follow our Helm Charts, and keep your original configuration: `TENANT_NAME`, `DOMAIN`, `APP_SECRET_TOKEN`, your database credentials, and your field-level encryption keys. Tines reads your tenant name and domain from the environment on startup. Your encryption keys are what make the restored credentials readable.

For a hosted database, set `db.create` to `false` and point `DATABASE_HOST` at your restored instance:

```yaml
db:
  create: false

tinesApp:
  databaseConfiguration:
    DATABASE_HOST: <restored-database-host>
    DATABASE_NAME: tines_production
    DATABASE_USERNAME: tines
    DATABASE_PASSWORD: <password>
```

```bash
helm install tines-app tines/tines \
  -n tines \
  -f values.yaml
```

The `tines-app-setup` job migrates your database and then seeds it. Seeding stops when your restored database already contains a tenant. Follow it with:

```bash
kubectl logs -n tines job/tines-app-setup
```

To move your existing release, change `DATABASE_HOST` and run `helm upgrade` with the same values file. Leave your deployments at zero until your cache is empty.

If the chart created your database, restore your backup into that database before `tines-app` and `tines-sidekiq` start.



### Clear or restore your cache, then start Tines

Valkey and Redis hold your sign-in sessions and queued jobs. Decide which case applies:

- If your Valkey or Redis cluster was also lost, skip this step. A new cache starts empty.
- If your cache is still running and you don't have a cache backup from the same time as your database backup, flush it. It holds sessions and jobs for data your restored database no longer has.
- If you have a cache backup taken at the same time as your database backup, restore that backup instead and point REDIS_URL at it. Skip the flush.

To flush, run:

```bash
kubectl exec -n tines <cache-pod> -- redis-cli -n <database-number> FLUSHDB
```

Start Tines again by scaling your tines-app and tines-sidekiq deployments back to their previous replica counts. Sign in and confirm that your stories, credentials, and cases are present. If you flushed or started with an empty cache, you sign in again and queued jobs from before the restore are gone.