GitOps: Argo CD and Flux
Typed panels for Argo CD Applications and Flux Kustomizations and HelmReleases, and a warning before any change a controller will undo.
Day-2 operations · New in v0.3.0
Change a cluster safely, and see the kubectl behind it
PodSteer can change a cluster as well as read it: apply a manifest of any kind, roll a workload back, drain a node, scale or restart many objects at once. Each write shows what it will do before it does it, and the kubectl command it is equivalent to.
Apply any kind, without destroying fields
The editor applies any kind, custom resources included. Declared intent — a pasted file, a new object, a duplicate — goes through server-side apply, so a manifest that does not mention a field leaves it alone. When another manager owns a field you are changing, PodSteer names the field and its owner and changes nothing until you have read who holds it.
Validation is a server-side dry run: your API server's own admission chain answers, and its words are shown rather than paraphrased.
What you can do
- Rollout history with a diff between revisions, and rollback, for Deployments, StatefulSets and DaemonSets.
- Cordon, uncordon and drain, with a preview of exactly which pods a drain will evict or refuse — the same plan it then executes.
- Evict through the PodDisruptionBudget-respecting Eviction API, not delete.
- Bulk delete, evict, restart, scale and cordon, with a review step; one refusal does not abort the rest.
- Set image, resize a pod or a sidecar container in place, run a CronJob now, suspend or resume it.
- Create from a skeleton, duplicate, and compare two objects side by side — across namespaces and across clusters.
- Edit one key of a Secret or ConfigMap, after it has been revealed on screen.
- Export any table to CSV in the columns you see; the pod table exports every pod matching the filter, not only the page on screen.
Guardrails you choose
Mark a group of clusters as production and destructive actions ask you to type the object's name. Mark one read-only and PodSteer refuses its own writes to it. The Scale dialog warns when an autoscaler owns the replica count, and every write to an object Argo CD or Flux manages says that the controller will put it back.
The kubectl behind every button
Every write dialog shows its kubectl equivalent, and any row can be copied as a kubectl command. The GUI teaches the CLI rather than hiding it.
Be honest
What it does not do.
- Read-only and production marks are guards against your own mistakes, set on your machine. They are not permissions: what an account can do is decided by the cluster's RBAC.
- There is no bulk drain. Draining several nodes at once can take a cluster down, and the per-node preview is the safety that would be lost.
- A rollback copies the target revision's pod template onto the live object; containers added since that revision are not removed.
- In-place pod resize needs Kubernetes 1.33 or later, and resizing a sidecar container needs 1.37. When a server accepts a sidecar resize but does not apply it, PodSteer says so.
Read next
Cluster assessment
Ranked findings instead of a pod list: crash loops, OOM kills, pods that will never schedule, sizing problems — each with what to do about it.
Shells and terminals
Exec, debug containers, node shells, an in-cluster shell and your own local terminal — with the pods PodSteer creates cleaned up after.