Skip to main content
The defaults are chosen so a first install works on any cluster. Several of them trade security for that. Work through this checklist before an installation holds production data. Each item names the setting and links to the page that explains it.

Access to the Admin UI and API

Publish the Admin UI over HTTPS. Don’t install with --quick-access or --no-ingress: both serve the Admin UI over plain HTTP. Set gateway.ingress with a hostname and a TLS certificate. See Production install.
Keep secure cookies on. Leave adminServer.env.cookieSecure unset, so cookies are Secure in production, and serve the Admin UI only over TLS.
Use a stable JWT secret from a Secret. Generate one of at least 32 characters with openssl rand -hex 32, and reference it with adminServer.secret.jwtExistingSecret rather than a Helm value. See Secrets.
Replace the bootstrap admin password. Change it at first sign-in, then delete the Secret that holds the generated password. If you set adminServer.bootstrap.adminUser.password yourself, keep mustChangePassword: true.
Protect the metrics endpoint. /metrics is served without authentication by default. Set adminServer.metrics.basicAuth.existingSecret, or username and password. The chart’s PodMonitor uses the same credentials.

Identity

Sign in through your identity provider. Configure Google or Microsoft Entra ID (Azure AD), and map groups to roles.
Turn off local sign-in once single sign-on works. Set adminServer.env.disableLocalAuth: "true". Keep the steps in Reset admin password for recovery.
Grant least privilege. Give each user the narrowest role in RBAC in the Admin UI. Running queries from the query console needs the clusters.query permission; write lets a query change data. If you don’t use the console, set adminServer.queryConsole.enabled: false.

Kubernetes boundaries

Enforce the admission policy. On Kubernetes 1.30 or later, make sure adminServer.rbac.admissionPolicy.validationActions is [Deny], so the policy refuses violations instead of only warning. See What the admission policy refuses.
Confine databases to a namespace prefix. Set adminServer.managedNamespaces.pattern, for example "falkordb-.*", and list any other namespaces to protect in adminServer.managedNamespaces.denied. See Database namespaces.
Turn on network policies. Set networkPolicy.enabled: true on a CNI that enforces them, set networkPolicy.ingressControllerNamespace to where your ingress controller runs, and set networkPolicy.ingressControllerPodSelector to its pod labels so nothing else in that namespace can reach the Admin Server. See Network policy.
Keep Pod Security Admission labels. The installer labels the release namespace enforce=restricted (baseline only while dm-sql-to-falkordb is enabled). Keep it there, and label the database namespaces at baseline. Don’t enable the bundled CSI drivers in production, because they need privileged. See Pod security.

Data

Turn on TLS for deployments. Choose TLS in the Encryption step when you create each deployment. It can’t be added later. See TLS.
Back up to object storage. Create an S3, GCS or Azure Blob repository, mark it as the default, and test a restore. The default repository is a volume in the same cluster. See Backups & schedules.
Back up the Admin Server’s volume. It holds the audit log and the license state. See Admin Server backup.
Keep credentials out of values files. Supply the license with adminServer.license.existingSecret, the Browser key with falkordb-browser.browser.encryption.existingSecret, and the registry credentials with an existing pull Secret. See Enterprise license and Private registries.

Supply chain and monitoring

Verify and pin images. Verify the chart and image signatures before you install, and pin images by digest. See Supply chain.
Turn on alerts. Set podMonitor.enabled and alertRules.enabled, so failed backups, broken replication and full volumes reach you. See Monitoring.
For how each control works and where its limits are, read Security.