Released
PodSteer v0.1.0
The first release anyone could download. It opened each cluster on an assessment rather than a list — capacity measured against what pods reserve, and pods explained rather than described — and it already held several clusters at once, one per tab.
What changed
The cluster overview
- Capacity measured against requests: CPU, memory, ephemeral storage and the pod slots a pod could actually land on.
- Node disks read from the kubelets, and persistent storage assessed.
- What is using the cluster, and when its Kubernetes version stopped being patched.
- Findings you can snooze per object, with an alert sound for each severity.
Pods, assessed
- Why a container restarted, not just how often — and the probe trap behind many restart loops.
- Stuck deletions explained, and what constrains a pod's scheduling.
- Replicas running different builds of the same image.
- A warning when a pod is close to its memory limit.
Working with a cluster
- A YAML editor in the side panel, with managedFields hidden until asked for.
- Logs that stay fast at any length, with matches marked in place.
- Port-forwards that survive the pod behind them being replaced.
- Secret values hidden until one is revealed, deliberately and briefly.
- A GitOps column saying when an object is reconciled from Git.
Clusters
- Several clusters open at once, one per tab.
- Add a cluster by merging a kubeconfig, and organise clusters into groups.
- An unreachable cluster is reported as an error, never as a green tick.