Skip to main content
You need to see what is really running in a cluster, and fix it, without switching kubectl contexts or handing out kubeconfigs. Every cluster connected to Ankra has a Kubernetes menu that lists the resources the Ankra Agent reports, and lets you act on them within your organisation permissions. This page covers what Ankra adds on top of the Kubernetes objects and where each one lives in the menu. For what a Deployment or a PersistentVolume is, see the Kubernetes documentation.

Where Everything Is

Open Infrastructure → Clusters → <your cluster> → Technical details → Kubernetes (/next/clusters/<cluster-id>/kubernetes). The side menu switches to the Kubernetes menu; the cluster’s name at its top takes you back to the cluster. Pod Disruption Budgets are under Workloads, and Network Policies under Access & policies. On All resources the access group is headed RBAC. Endpoints, LimitRanges, CSINodes and VolumeAttachments have no menu entry; read them with the CLI, which takes any kind:

Resource Lists

Each list has a name search, a Resource type picker that jumps to another kind, and, for namespaced kinds, a Namespace field: type a namespace and click Apply namespace, or leave it on All namespaces. Click a column heading to sort. Long lists page with Next page and First page. Above the list a line says how fresh the data is, for example Showing resources from the platform cache. Last sync was 12 seconds ago. When the agent is not reporting it reads The cluster is offline. These are previously recorded resources. or These resources may be out of date., and you see the last copy Ankra recorded. Refresh reads the list again. To delete several resources at once, open Select resources for deletion under the list, tick up to 20 rows on the page, and click Review selected resources. The review that follows is the same as for one resource, below.

What Every Resource Page Has

Click a resource to open its page. The tabs are: A namespace page also has Browse this namespace, with a link to each namespaced kind filtered to that namespace. The cluster-wide event list is Cluster → Events. On the YAML tab, values that look like credentials, and every Secret value, are shown as [protected].

Change a resource

Edits, restarts, deletes and Secret reveals all start from Review resource actions on the resource page. Ankra reads the object live from the cluster, then opens Review <resource name> with the actions your role allows: Every action asks you to type the resource name in Type <resource name> to confirm before the confirm button turns on. When the change is accepted the dialog shows <resource name>: request accepted; check the result on the resource’s Overview or Events tab, since an accepted request is not proof that the workload is healthy. Review resource actions only appears while the list is live. On cached or offline data there is nothing to review. A resource a Stack deployed is marked managed by <kind> in the review, with an Edit owning stack link. Ankra offers Restart on it, but not Edit manifest or Delete: change it in its Stack, which is the source of truth. A direct change is refused with This object is managed. Change its source of truth. If your role cannot act, the review says Your role cannot perform this action. See Roles & Access for which role holds which permission.

Workloads

Workloads holds the workload kinds and Pods. The pod list has the same name search and namespace filter, sorts by name, namespace, phase, restarts, ready containers or age, and each row has View logs and a menu (the three dots). On a pod that is not running, has unready containers or has restarted, the menu has one entry worded for what the row shows: Explain <phase> status with Ankra, Investigate unready containers with Ankra or Investigate restarts with Ankra. Choosing it asks Ankra AI at once why the pod is in that state. If you can read security findings, the menu also has View security findings for the pod. A pod page has Overview, Logs, Metrics, Exec and YAML tabs. Overview shows the phase, containers ready, restarts, each container’s state and requests and limits, the pod’s conditions and its Events, newest first. Delete pod in the page header deletes the pod after you type its name; the dialog says whether its controller recreates it. The pod-level tools have their own pages:
  • Logs - follow a container’s output, or search the cluster’s logs.
  • Pod Terminal - a recorded shell in a running container, on the Exec tab.
  • Debug Pods - a pod that impersonates this one under an image with the tools you need.
  • Cluster Metrics - CPU and memory for nodes and pods.

Networking

Networking holds Services, Ingresses and Ingress Classes. A Service page has a Pods tab listing the pods its selector matches, which is the quickest way to see why a Service has no backends. The Services list shows each Service’s type, cluster IP and ports; the Ingresses list shows each Ingress’s hosts and ingress class. Clusters Ankra creates can include a networking stack at create time: Traefik as the ingress controller, cert-manager and a Let’s Encrypt issuer. Network Policies are under Access & policies; Cluster Security scores how tightly they are scoped.

Storage

Storage holds Persistent Volumes, Persistent Volume Claims, Storage Classes and CSI Drivers. The Persistent Volumes and Persistent Volume Claims lists show capacity, storage class and status, and the Storage Classes list shows each class’s provisioner, Reclaim policy and volume binding mode.
Changing a Stack can delete volume data. Removing a manifest from a Stack, or renaming one, makes Ankra delete the objects in it, PersistentVolumeClaims included. Ankra keeps an object that another live manifest or application on the cluster also declares, or that the Git repository’s head declares under another Stack; an object only this manifest declares is deleted. With a Delete reclaim policy, the default for most CSI drivers, the disk behind a deleted claim is destroyed too. Read Protect Persistent Data Before Changing a Stack before you rename or remove a manifest that holds a namespace or a claim.

Configuration

Configuration holds ConfigMaps and Secrets. The ConfigMaps list shows each one’s key count and whether it is immutable; the Secrets list shows type, key count and immutability, never values.

Secrets

A Secret’s values stay hidden. On its YAML tab every value reads [protected]. To see the values:
  1. Open the Secret and click Review resource actions.
  2. Click Reveal secret values, type the Secret’s name and click Reveal secret.
The values appear in their Kubernetes form (base64 under data). They hide again after 30 seconds, when the browser window loses focus, or when you click Hide values. Reveal secret values is only offered when your role holds the kubernetes.secrets_reveal permission, which the owner, admin and operator roles hold; member and viewer see key names only. Each reveal is checked and written to the audit log. From the terminal, ankra cluster get secrets shows key names and a digest for each value. A digest (hmac-sha256: and 32 characters) changes when its value changes. It is keyed per cluster with a key only the platform holds, so it cannot be used to check a guessed value, and the same value on two clusters shows two different digests. To read one Secret’s values, name it, give its namespace and add --reveal:
--reveal needs the same kubernetes.secrets_reveal permission, and each reveal is written to the audit log like one from the portal. It reads exactly one Secret, so it is refused on a listing: without a name, with -A or with -l. Values under data stay base64-encoded, as kubectl prints them. The command exits 7 when your role lacks the permission, 3 when the Secret does not exist, 1 when the platform cannot read live values, for example because the cluster is not connected, and 2 for a reveal on a listing. To keep secrets out of plain text in Git, see SOPS Secrets. For values shared across clusters and stacks, use Variables instead of hand-edited ConfigMaps.

Access & policies

Access & policies shows the Kubernetes RBAC objects inside the cluster: Roles and Cluster Roles with their rules, their bindings with role and subjects, Service Accounts, and Network Policies. This is not where you give a teammate access. Ankra has its own access control:
  • Roles & Access decides what each member can do in Ankra, per organisation, cluster or cluster group.
  • Cluster Access grants a member kubectl access to a cluster. Ankra turns each grant into a RoleBinding or ClusterRoleBinding, which then appears in this menu.

Cluster and Helm

  • Nodes - the Kubernetes Node objects. A node page shows its Conditions, the Workloads on it and Scheduling, with Review cordon and Review node drain. See Kubernetes Nodes. To add, remove or resize the machines themselves, see Cluster Nodes.
  • Namespaces - each namespace’s status; a namespace page lists its pods and links to every kind inside it.
  • Events - every Kubernetes event in the cluster, with reason, object and message.
  • Custom resource definitions - every CustomResourceDefinition, with its kind, group and scope.
  • Helm releases - every Helm release, with chart, revision and status. A release page has Overview, Resources (the objects the release declares) and History (the latest 25 revisions) tabs, and three buttons: Each asks you to type the release name. A release an Ankra add-on or Stack owns is changed in its Stack, because the Stack would re-apply its own version: the upgrade dialog links Review changes in the owning stack, and the rollback and uninstall dialogs say This release belongs to an Ankra stack. See the Add-on catalog. A release whose values hold passwords, tokens or keys says This release contains protected values. Use its source of truth to preserve those values during an upgrade. and offers no values editor.

Next Steps

Start from a problem: open Workloads → Pods, sort by Most restarts, and choose Investigate restarts with Ankra on the pod that keeps restarting. Or open the pod, check Events on its Overview tab, then its Logs. Press ⌘J (Ctrl+J) on any page to open Ask Ankra about the cluster you are on. For the full order of checks when a cluster misbehaves, see Troubleshoot a Cluster.