DEV Community

jeffrey
jeffrey

Posted on

CockroachDB fingerprint matches need secure and insecure modes separated

CockroachDB fingerprint matches need secure and insecure modes separated

CockroachDB can run with TLS and authenticated client certificates, or in an explicitly insecure local mode with no network encryption or authentication. A product fingerprint can find database nodes and management planes, but it cannot decide which mode a particular cluster uses.

On 2026-11-21 at 04:47 CST, app="CockroachDB" returned 5,767 ZoomEye matches. title="CockroachDB" returned 99 with the same sub_type=all, page 1, and page size 1 settings. The application fingerprint is more specific than the title, yet the count still describes index matches rather than confirmed insecure clusters.

This is a case where comparing the two queries changes the interpretation. A title-led search would miss most application matches. An application-led search still needs protocol-level verification because an application fingerprint can identify a node or console without proving how it authenticates SQL clients.

Inspect the database and DB Console separately

The SQL listener and the DB Console have different users, certificates, and network paths. A secure SQL port does not prove that the console uses a strong identity model. An exposed console does not prove that SQL clients can connect without certificates.

Inventory the SQL ports, HTTP admin ports, service certificates, client certificate authorities, and console SSO configuration. Check whether the cluster was started in insecure mode and whether that setting is still present in unit files, container arguments, or orchestration manifests.

The official local-cluster guide labels the insecure mode clearly and states that it has no network encryption or authentication. Use it for bounded local testing only. Production deployments should follow the secure deployment path and verify that nodes, clients, and the console use the intended trust roots.

Test authentication from the client side

Connect with an authorized client certificate and a role that has the required privileges. Then try the same operation without a valid certificate and from a certificate mapped to a less privileged identity. Confirm that SQL authorization limits the operation even after TLS authentication succeeds.

Test the DB Console separately. A role that can view cluster health should not automatically administer SQL users or certificates. Review OIDC or other SSO mappings if enabled, and check session revocation when an external identity loses access.

Verify the network boundary

Restrict database and console ports to the networks that need them. Do not expose the insecure mode as a shortcut for debugging. If a monitoring system needs metrics, give it a narrow endpoint and credential rather than direct console access.

The useful finding is a verified cluster mode, a known console identity path, and a tested SQL role. The 5,767 application matches are a discovery signal, not an insecurity count.

References

Top comments (0)