periscopev1.0.7

docsusagestorage

Storage (PVs, PVCs, StorageClasses)

The **STORAGE** sidebar group surfaces the three K8s storage objects operators reach for during incidents — `PersistentVolumes`, `PersistentVolumeClaims`, and `StorageClasses`. Live-watched lists with detail panes tha…

The STORAGE sidebar group surfaces the three K8s storage objects operators reach for during incidents — PersistentVolumes, PersistentVolumeClaims, and StorageClasses. Live-watched lists with detail panes that pair the Bound PV ↔ PVC for a single click.

PersistentVolumes

PersistentVolumes list with the periscope-audit-bound PV selected — Bound, gp3, 5Gi, ebs.csi.aws.com source
PersistentVolumes list with the periscope-audit-bound PV selected — Bound, gp3, 5Gi, ebs.csi.aws.com source

List

The page header shows the PV count and the status filter pills — all / available / bound / released / failed. Each row carries:

  • NAME — metadata.name. Dynamically-provisioned PVs use the pvc-<uid> convention.
  • STATUS — colored chip (Bound green, Available neutral, Released/Failed warning).
  • CAPACITY — spec.capacity.storage.
  • CLASS — spec.storageClassName.
  • RECLAIM — spec.persistentVolumeReclaimPolicy. Retain keeps the underlying volume on PVC delete; Delete reclaims it.
  • AGE — relative timestamp.

Detail (describe)

The detail pane's describe tab carries:

  • Header — name + the cluster-scoped chip (PVs are cluster-scoped) + edit / labels / delete actions.
  • Runtime cards — STATUS / CAPACITY / AGE.
  • Storage class / Access modes / Reclaim policy / Volume mode — the four config knobs, each one row.
  • Source — the CSI driver name (csi.ebs.csi.aws.com in the screencap, indicating the AWS EBS CSI driver provisioned this PV). In-tree drivers (legacy) render as awsElasticBlockStore, gcePersistentDisk, etc.; the source row is the easiest way to spot in-tree usage you should migrate off.
  • Bound claim — namespace/name of the PVC currently bound to this PV. Click to jump to the PVC detail.
  • Annotations — provisioner-set annotations are surfaced verbatim (e.g. pv.kubernetes.io/provisioned-by, the deletion-secret annotations).

The screencap above is the EBS-backed PV that holds Periscope's own audit log on peri-server — a 5Gi gp3 RWO volume bound to periscope/periscope-audit.

Detail (yaml + events)

The yaml tab shows the canonical manifest including the source spec block (csi: { driver, volumeHandle, fsType, volumeAttributes }) that the describe tab summarizes. The events tab is the per-PV event feed — most operationally interesting for surfacing provisioner failures (ProvisioningFailed, Resizing).

PersistentVolumeClaims

PersistentVolumeClaims list with the periscope-audit PVC selected — Bound, gp3, 5Gi, bound to a pvc-uid PV
PersistentVolumeClaims list with the periscope-audit PVC selected — Bound, gp3, 5Gi, bound to a pvc-uid PV

List

Same shape as the PV list, scoped per-namespace via the namespace picker. Status filter pills — all / bound / pending / lost. Each row carries:

  • NAME / NAMESPACE — the claim itself.
  • STATUS — colored chip (Bound green, Pending amber, Lost warning).
  • CAPACITY — status.capacity.storage (the actually-bound size).
  • CLASS — spec.storageClassName.
  • AGE — relative timestamp.

Detail (describe)

  • Runtime cards — STATUS / CAPACITY / AGE.
  • Storage class / Access modes — same as the PV detail.
  • Bound PV — pvc-<uid> link to the matching PV detail page. The pair claim → bound PV is one click each direction; useful when a PVC is Pending and you need to see whether a matching PV exists.
  • LABELS / ANNOTATIONS — Helm-managed labels render verbatim (the screencap shows the standard app.kubernetes.io/* set); provisioner annotations include volume.kubernetes.io/storage-provisioner=ebs.csi.aws.com and volume.kubernetes.io/selected-node=ip-… (CSI external-provisioner state).

Common Pending-state debugging

When a PVC stays Pending, the events tab surfaces the ProvisioningFailed reason verbatim. Common culprits:

  • No matching StorageClass. PVC requests storageClassName: foo but no SC named foo exists. Events tab shows storageclass.storage.k8s.io "foo" not found.
  • Provisioner can't allocate. AZ mismatch (PVC on a node in us-west-2c but the EBS CSI driver IRSA only allows us-west-2a), quota exhaustion, missing IAM permission. Events tab carries the CSI driver's verbatim error message.
  • WaitForFirstConsumer. PVC is Pending deliberately — the StorageClass uses WaitForFirstConsumer topology and no Pod has been scheduled to a node yet. Not a problem; resolves when a Pod references the PVC.

StorageClasses

The StorageClasses page lists every storage.k8s.io/v1 StorageClass on the cluster. Operationally simpler than PV/PVC — one row per SC with PROVISIONER, RECLAIM, VOLUME BINDING MODE (Immediate / WaitForFirstConsumer), and DEFAULT chip (set when the SC carries the storageclass.kubernetes.io/is-default-class=true annotation).

The detail pane shows the full parameters block (the per-driver config — for EBS, the type, encrypted, iopsPerGB, etc.) plus labels / annotations.

→

No screenshot for this page — the layout matches the PV list shape with a smaller column set.

Live updates

All three list pages are wired to the cluster's watch streams. PVC binding state transitions (Pending → Bound) propagate without a refresh — useful for staring at a fresh PVC and watching the provisioner pick it up.

RBAC

The standard view ClusterRole grants get + list on PVs, PVCs, and StorageClasses cluster-wide. Role-restricted users see only the PVCs in namespaces their Role permits; PVs and StorageClasses are cluster-scoped, so a denial there hides the entire page.

Related docs

  • workloads.md — Pods reference their PVCs in spec.volumes; the workload detail surfaces the volume mounts.
  • ../setup/watch-streams.md — the SSE plumbing that makes these list pages live.
  • cluster-overview.md — the Storage card on the overview page summarizes total/bound capacity.
← previousNetworkPoliciesnext →Custom Resources (CRDs)
edit this page on github →