Viewing and changing resources
Every resource type opens as a table, and every row opens a detail drawer. You can edit, create, and act on objects from there. Whatever you do runs with your own credential, and the API server has the final say. For the model behind this, see Operations and security.
View objects
Open a type from the sidebar or the command bar (:deploy, :svc, :pvc). Built-in kinds have purpose-built columns and status logic: a Pod's status reason follows kubectl get pods (CrashLoopBackOff, OOMKilled, Terminating), a Deployment shows ready, up-to-date, and available counts, and Services show type, ports, and ready endpoints.
Custom resources and everything else get a generic view built from the type's printer columns, with a status tag when the type has a Ready, Status, or Phase column.
Cluster-scoped and multi-cluster lists
With several clusters selected, a table merges them and adds a cluster column. Sorting, filtering, and counts cover the merged set.
The detail drawer
Press Enter on a row, or click it, to open the drawer. The URL changes with it, so the drawer is linkable. On a phone it is a full page.
| Tab | What it shows |
|---|---|
| Overview | Status, conditions, key fields, owners, and, for managed objects, a link to the Flux or Helm source |
| Logs | For pods and workloads. See Operations |
| Shell | For pods. See Operations |
| Inventory | For Flux Kustomizations: the objects they manage. See Flux |
| Release | For Flux HelmReleases: the Helm release underneath. See Flux |
| Report | For Trivy, Kyverno, and other scanner report objects: the report's rows (CVEs, checks, results) |
| Security | Kyverno, Trivy, and Kubescape findings for this workload. See Security |
| Flows | For pods, workloads, Services, and namespaces, when Hubble is detected. See Network |
| YAML | The live object. Read, copy, or edit it |
| Events | All events for this object, newest first |
| Related | What owns this object and what it owns |
The Events tab shows every event for the object, Normal and Warning, newest first, with the component that reported each one. It loads when you open the tab and stays live while it is open. Events of cluster-scoped objects (nodes, volumes, namespaces) come from the default namespace. Kubernetes keeps events for about an hour, so older ones are gone, and an object recreated with the same name shows the earlier one's events until then. Event objects themselves have no Events tab.
Tabs appear in that order, and only those that apply to the type and, for Security and Flows, to the clusters you selected. Use the arrow keys, Home, and End to move between tabs. l, s, e, and d jump to Logs, Shell, YAML (in edit mode), and Overview from a table. See Navigation.
Edit YAML
- Open the YAML tab and choose Edit, or press
eon a row. - Change the document.
- Choose Review changes. v8s runs a server-side dry run against the live object (
dryRun=All), so admission webhooks such as Kyverno or Gatekeeper reject a bad change before you commit it. You see a diff against the live object, plus any warnings or errors from the server. - Choose Apply once the dry run passes.
Applying is a normal update of the object you opened, like kubectl edit. Only fields you changed become yours, and deleting a field really deletes it.
Large documents
Documents over 1 MB open read-only, and so do objects deleted on the server.
Who else manages this field
During review, v8s compares the changed paths with the object's managedFields and names other managers:
spec.replicasis managed bykustomize-controller. Flux will revert this change.
Read that before you apply. A change to a Flux- or Helm-managed field is overwritten at the next reconcile.
GitOps banner
If the object carries Flux ownership labels (kustomize.toolkit.fluxcd.io/name, helm.toolkit.fluxcd.io/name) or Helm's app.kubernetes.io/managed-by: Helm, the review step shows a banner naming the owner ("may revert this change. Change it at the source, or suspend it first"). The Overview tab of a managed object has an Open source link to the Kustomization or HelmRelease, which is usually where the change belongs. See Flux and Helm.
Conflicts
Your edit carries the version you opened. If the object changed on the server in the meantime, v8s tells you: while you edit, you can Reload (discard my edits) or keep editing. If you apply and the server reports a conflict, v8s shows what changed on the server since you opened the object and your changes against the live object, and offers Discard my edits and reload or Overwrite with mine. Nothing is overwritten without that second choice.
Server rejected it
A dry run that fails shows the server's message in place (invalid field, admission denial, quota, forbidden). Nothing was changed. Fix it and review again.
Create objects
Choose Create on any list page.
- From a template. Pick Namespace, ConfigMap, Secret, Deployment, Service, Ingress, CronJob, or Job. v8s fills a minimal manifest for the current namespace.
- From YAML. Paste your own manifest in the editor. Multi-document YAML is supported.
- Dry run first. Dry run checks the manifest against the cluster without creating anything. Create applies it.
Creation uses server-side apply with force=false, so a conflict with another field manager is reported, not forced over.
Actions
Actions appear on a row's action bar and in the bulk bar for a selection. They are disabled, with a reason, when your RBAC does not allow them. If your cluster has a linked write context, the tooltip says the action will run through it.
| Type | Actions |
|---|---|
| Deployment | Scale, Restart, Roll back |
| StatefulSet | Scale, Restart |
| DaemonSet | Restart |
| ReplicaSet | Scale |
| Node | Cordon, Uncordon, Drain (see Operations) |
| CronJob | Suspend, Resume, Trigger now |
| Flux and Helm objects | See Flux and Helm |
| Any type | Delete |
- Scale asks for the replica count.
- Restart patches the pod template's
kubectl.kubernetes.io/restartedAtannotation, the same thingkubectl rollout restartdoes. - Roll back lists the Deployment's previous revisions and rolls back to the one you choose.
- Trigger now creates a Job from the CronJob's template.
- Suspend and Resume set the CronJob's
spec.suspend.
These send small targeted patches that touch only their own field, so they do not overwrite concurrent changes.
Confirmations
Every destructive action asks first. How hard it asks depends on the blast radius:
| Situation | Confirmation |
|---|---|
| One object in a non-production cluster | A confirm dialog |
| More than one object, or objects in more than one cluster | Type the count to confirm |
| Any object in a cluster that belongs to a production group | Type the object name (or the count) to confirm |
Bulk dialogs list the exact objects and clusters. A selection always resolves to the objects you chose when you press confirm, even if the list moved underneath you. Retrying after a reconnect cannot apply an action twice.
On a phone, destructive actions use hold-to-confirm. See On a phone.
Secrets
Secret lists never download values: v8s watches Secrets as metadata only, so you see names, types, and ages. Opening a Secret does not fetch it either. The Overview tab says so, and the YAML tab loads the masked manifest on demand, tagged values masked.
Revealing is for the whole object, not per key. Reveal all fetches the Secret again with every value (base64-encoded, as stored) and tags the pane values revealed. It needs the get permission on that Secret. Mask again drops the values, and they are also dropped when you close the drawer, change tab, or open another object. You cannot reveal or mask while an edit is open.
Secret values are never written to logs or the recent activity list.
Custom resources
CRDs are discovered per cluster. They appear under Custom resources in the sidebar, grouped by API group, and complete in the command bar (:certificates, :ks). A CRD's list uses its own printer columns. Flux, Kyverno, Trivy, Kubescape, and Cilium types have dedicated views instead; see their pages. The CustomResourceDefinition list shows group, versions, scope, and a jump to its instances.