Slow incident triage
Operators can start from grouped health signals instead of jumping between raw object tables, terminals, and ad hoc commands.
Home / BeaverDeck
BeaverDeck
A self-hosted Kubernetes operations workspace for Cluster Insights, resource inspection, logs, exec sessions, and common day-2 actions. GPU visibility and capacity signals help teams understand how expensive compute is used by AI workloads.
Problems It Solves
BeaverDeck is built for teams that need to shorten the path from a cluster signal to a controlled action.
Scattered day-2 work
Logs, manifests, exec sessions, YAML apply, scaling, restarts, drains, deletes, and RBAC-aware actions live in one workspace.
Expensive compute blind spots
GPU visibility and capacity signals help teams understand AI workload placement and reduce waste on costly infrastructure.
Features
Inspect Kubernetes state, triage findings, and carry out common day-2 operations from one self-hosted workspace.
Cluster Insights
Review grouped findings for nodes, workloads, GPU capacity, networking, storage, security, and configuration.
Resource inspection
Browse common Kubernetes resources, events, manifests, and relationships without switching between separate tools.
Logs and exec
Stream workload logs and open terminal sessions, including explicit container selection for multi-container pods.
Controlled operations
Apply YAML, edit manifests, scale or restart workloads, and perform node or resource actions under BeaverDeck permissions.
GPU visibility
Surface allocation pressure, scheduling constraints, idle allocations, fragmentation, quotas, and workload placement signals.
Self-hosted access
Run BeaverDeck in your cluster with local users or configured Google, OIDC, and Microsoft Entra ID authentication.
Product Direction
BeaverDeck stays focused on practical workflows, clear resource state, and predictable ownership.
Operations visibility
Start from useful signals, then get to the resource, log, manifest, or action behind the signal.
Compute efficiency
Improve understanding of costly infrastructure, including GPU capacity used by AI workloads.
Self-hosted control
Keep deployment, access control, and operational configuration inside infrastructure you operate.
Insights-First Triage
Start from grouped signals, then open the resource, log, manifest, or operation connected to the finding.
Nodes
Readiness, pressure conditions, metrics availability, GPU visibility, and compute capacity signals.
Workloads
Pod and controller health, restart patterns, pending pods, resource pressure, and security context warnings.
Networking
Ingress and service signals that help narrow traffic-routing issues.
Storage
PVC binding state, volume usage, storage class visibility, and persistent storage pressure.
Operations Workspace
BeaverDeck keeps the confirmation and action path close to the signal.
Inspect and troubleshoot
- Browse pods, workloads, nodes, services, ingresses, config maps, secrets, PVCs, PVs, storage classes, CRDs, and events
- Inspect manifests as YAML
- Stream pod and workload logs
- Open exec sessions into running pods
Operate with control
- Apply YAML and edit resources through server-side apply
- Scale, restart, delete, evict, drain, and uncordon from the UI
- Track GPU visibility and capacity signals that affect AI workload placement
- Keep actions controlled with users, roles, and namespace-scoped permissions
Deployment
Install with the official Helm chart.
BeaverDeck stores auth configuration in a Kubernetes Secret. DATA_DIR is used only for
non-auth runtime metadata, such as update-check status.
helm upgrade --install beaverdeck oci://ghcr.io/arequs/charts/beaverdeck \
--namespace beaverdeck \
--create-namespace \
--set clusterName=your-cluster-name