
Loading

Loading
We use strictly necessary cookies to run this site, and analytics cookies to understand how it's used. See our Privacy Policy for details.

Separating etcd Client and Peer Trust Is a Control-Plane Decision etcd is often described as a key-value store, but in Kubernetes and similar systems it holds the state that everything else depends on.
Summary
A TLS listener that accepts both cluster peers and application clients may look simpler than separate transports. It also collapses two different trust decisions into one certificate population. The etcd transport security guide states that client-to-server and peer-to-server communication can both use TLS and client certificates.
RBAC and authentication are not enabled by default. That default is convenient for development and dangerous as an assumption in production. The same guide documents peer-client certificate authentication as a control against forged peers, while the configuration reference recommends enabling it.
Client certificates can carry identity. In etcd, when client certificate authentication is enabled, the certificate Common Name can be used as the etcd user. That makes certificate issuance, naming, renewal, and revocation part of authorization.
A certificate with an unintended CN may also map to a user with roles. Review the mapping before issuing a broad internal CA certificate to every workload. Separate trust domains when the deployment permits it.
Use a peer CA for cluster members and a client CA for applications and operators. Restrict peer certificate names and SANs, and make sure a client certificate cannot authenticate as a peer. Keep the peer listener on a private cluster network.
Do not expose it through an ingress controller or a broad service load balancer. RBAC needs its own verification. The etcd documentation notes that the /metrics and /health endpoints use a different handler and are not protected by V3 RBAC authentication.
That does not make every unprotected endpoint operational, but it means a successful etcdctl auth status check is not a complete exposure assessment. Inventory listeners, health endpoints, monitoring paths, and any gateway that forwards to them. Certificate rotation is a recurring control.
Confirm how client certificates are issued, how long they live, and what happens after revocation. Check whether applications fail closed when the etcd endpoint rejects them. A workload that falls back to plaintext or a stale credential can erase the benefit of the transport design.
Test the trust boundary from each network zone. A client certificate should fail against the peer listener. A peer certificate should fail against the client listener unless it is intentionally authorized.
A connection without a client certificate should fail when client authentication is required. Use disposable test credentials and record the result. The design should also identify who can change etcd configuration.
A role that can modify the datastore may be able to alter Kubernetes objects, authentication configuration, or workload definitions. Treat write access as control-plane authority. Log administrative sessions, use a dedicated management network, and keep backups protected with the same seriousness as the live data.
The central point is not that every etcd installation needs a complex PKI. It is that client and peer authentication answer different questions. One asks whether an application or operator may read or write keys.
The other asks whether a node is allowed to participate in the cluster. Combining those questions removes a useful separation that is hard to reconstruct after an incident.
KazaSec's take
Phishing and credential-based attacks succeed because they target people, not just infrastructure, technical controls like SPF, DKIM, and DMARC only ever close part of that gap. The rest comes down to whether a team can actually spot the fake, and whether a compromised credential can still be reused anywhere else.
Coverage details
We've archived 754 other articles touching the same topic (cybersecurity) , see the full security news archive.
Relevant from KazaSec
More coverage on this topic
We help organizations find and fix the gaps before they make headlines.