CacheDB 1.9 security

Make the deployment boundary explicit.

CacheDB has documented security controls and documented transport limitations. A secure deployment is a system design: trusted network placement, edge controls, identity isolation, and operational access all matter.

Read before production

Do not assume encryption is supplied by the proxy.

CacheDB 1.9 native PostgreSQL and MySQL proxy paths do not support TLS on either leg. PostgreSQL, Oracle, and DB2 translation backends do not negotiate TLS. Redis connections—including cache, metrics, and Sentinel—do not support TLS. Place these components on a trusted private network or provide encryption at the deployment layer.

Current 1.9 boundaries

Known limits are part of the product contract.

Design reviews should treat the following as deployment constraints, not as future configuration choices.

Native proxy transport

PostgreSQL and MySQL native proxy traffic is plaintext between client, proxy, and backend.

Translation paths

PostgreSQL, Oracle, and DB2 backends do not negotiate TLS. SQL Server backend TLS currently accepts any certificate.

Redis transport

Cache, metrics, and Sentinel connections have no TLS support in this release.

Console edge

The console does not terminate TLS or provide built-in login rate limiting.

Controls around the product

Build the trusted edge deliberately.

Use the network and identity layers to match the threat model of the deployment.

Private network placement

Keep proxy, database, and Redis on a trusted segment. Use a service mesh, TLS sidecar, or network overlay where encryption in transit is required.

Console protection

Place externally reachable console access behind a reverse proxy or WAF for TLS, throttling, and network access control.

Identity isolation

Do not share a cache namespace between service identities with different data permissions or row-level visibility.

Operational controls

Authentication, audit, and policy changes remain reviewable.

The console supports local authentication with Argon2id password hashing and optional OIDC. Roles separate viewing, operations, and administration. Console sessions use HttpOnly and SameSite=Strict cookies, and protected mutations require CSRF protection.

Protected mutations and authentication or authorization failures are recorded with actor, action, target, outcome, and peer context. Cache-policy scripts execute in a restricted Rhai engine, but their operational effect still requires review and change control.

Read console operationsRead policy scripting boundaries

Deployment review

Bring the threat model into the evaluation.

We can review topology, encryption requirements, database identities, and console access controls against the documented 1.9 boundaries.