Dashboard

The dashboard is the browser view of the same controller API that kctl and Terraform use. It runs inside the controller process. There is no second service to install. Anything you can see or change there, you can also do from the command line, and every page names the matching kctl command and Terraform resource.

Open it

The controller serves the console over HTTPS on web.listenAddr, by default 0.0.0.0:9443. Open https://<controller>:9443/. The NixOS module turns it on with services.kcore.dashboard.enable and sets the address with services.kcore.dashboard.listenAddress. A controller run by hand needs web.enabled: true in controller.yaml.

Sign in

An operator signs in with a password and a TOTP code. There is no shared admin account and no SSO yet (no OIDC, SAML, or WebAuthn).

  1. A cluster-admin sets the first password with a certificate: kctl operator set-password alice.
  2. On first sign-in the dashboard shows a QR code. Scan it with an authenticator app and enter the first code. The dashboard then shows ten recovery codes, once.
  3. After that, every sign-in asks for the password, then a six-digit code or one recovery code.

If a phone is lost, kctl operator reset-2fa alice clears TOTP and the recovery codes and ends the operator’s web sessions. After maxFailedLogins wrong attempts in a row (default 5) the operator is locked for lockoutSecs (default 900). Sessions last 12 hours with a one-hour idle limit by default.

The Account page changes the password and sets up TOTP. The API tokens page creates tokens for scripts. A token is shown once, lasts 1 to 365 days (90 by default), can be capped below the operator’s role, and does not ask for a second factor. It cannot create tokens or change the password.

Roles and audit

The dashboard calls the controller as the signed-in operator, so the same roles apply: read-only, vm-admin, cluster-admin. The controller refuses what a role may not do. Mutating calls and console opens are written to the append-only audit log, which has its own page and a kctl get audit-events twin.

What is in it

GroupPages
OverviewCluster summary with charts.
ComputeVirtual machines, workloads, containers, operations (live migrations, drains, failovers in flight).
NetworkNetworks, host networking, security groups.
StorageVolumes, snapshots, snapshot policies, disk layouts, host storage, storage classes, Ceph clusters, shared filesystems, object stores.
IngressGateways, routes, WAF policies, WAF events.
ClusterNodes, Kubernetes, cluster updates, release, cluster, high availability events, replication, replication conflicts.
IdentityOperators, API tokens, my sessions, SSH keys, certificates, revocation list, PKI.
ServicesSlurm, PostgreSQL.
ObserveGPUs, cluster health, compliance, audit log, crypto configuration, SBOM.

Lists filter on any column, shown or hidden. A row opens a drawer with the full record, and the same drawer carries the actions for that object: start and stop a VM, migrate it, drain a node, approve a node, create or delete a resource from a form, and plan a change first where the API has a plan call.

Things the dashboard does well

Limits