Homelabv4/infrastructure/vault/manifests/raft-snapshot-cronjob.yaml

---
# Periodic Vault raft snapshot, shipped to the VPS MinIO backup receiver.
# This is the DR path for Vault's data independent of the unseal-key custody
# story - a Ceph/cluster-loss disaster is recovered by standing up a fresh
# Vault and `vault operator raft snapshot restore`ing the latest one of these,
# not by anything to do with the unseal key itself.
#
# Requires a one-time manual step after VPS bootstrap: store the VPS MinIO
# root credentials (see infrastructure/vps-standby/minio/values.yaml) into
# this cluster's Vault so ESO can hand them to the CronJob:
#   vault kv put secret/vps-minio-credentials \
#     access-key=<minio-root-user> secret-key=<minio-root-password>
apiVersion: external-secrets.io/v1
kind: ExternalSecret
metadata:
  name: vps-minio-credentials
  namespace: vault
spec:
  refreshInterval: 1h
  secretStoreRef:
    kind: ClusterSecretStore
    name: vault-backend
  target:
    name: vps-minio-credentials
    creationPolicy: Owner
  data:
    - secretKey: access-key
      remoteRef:
        key: vps-minio-credentials
        property: access-key
    - secretKey: secret-key
      remoteRef:
        key: vps-minio-credentials
        property: secret-key
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: vault-raft-snapshot-script
  namespace: vault
data:
  snapshot.sh: |
    #!/bin/bash
    set -euo pipefail

    # VPS's Netbird address - replace with the actual peer IP/hostname once
    # the VPS is bootstrapped and joined to the mesh (scripts/vps-bootstrap.sh).
    VPS_MINIO_ENDPOINT="${VPS_MINIO_ENDPOINT:-vps-minio.netbird.internal:30900}"
    BUCKET="vault-raft-snapshots"
    SNAP_NAME="vault-raft-$(date -u +%Y%m%dT%H%M%SZ).snap"

    ROOT_TOKEN="$(kubectl -n vault get secret vault-init-keys -o jsonpath='{.data.VAULT_ROOT_TOKEN}' | base64 -d)"

    # `vault operator raft snapshot save` fails against a standby node with
    # "incomplete snapshot, unable to read SHA256SUMS.sealed file" - it must
    # run directly against the active HA leader, not a proxied standby.
    # Found 2026-08-17: this job had been failing every run since it was
    # added, hardcoded to vault-0 (frequently a standby). Detect the actual
    # leader at runtime instead of hardcoding a pod name.
    echo "==> Finding the active Vault leader..."
    LEADER_POD=""
    for pod in vault-0 vault-1 vault-2; do
      is_leader="$(kubectl -n vault exec "$pod" -- env VAULT_ADDR=http://127.0.0.1:8200 VAULT_TOKEN="$ROOT_TOKEN" \
        vault status -format=json 2>/dev/null | jq -r '.is_self // false')"
      if [ "$is_leader" = "true" ]; then
        LEADER_POD="$pod"
        break
      fi
    done
    if [ -z "$LEADER_POD" ]; then
      echo "ERROR: could not find an active Vault leader among vault-0/1/2" >&2
      exit 1
    fi
    echo "==> Taking raft snapshot from ${LEADER_POD} (active leader)..."
    kubectl -n vault exec "$LEADER_POD" -- env VAULT_TOKEN="$ROOT_TOKEN" \
      vault operator raft snapshot save "/tmp/${SNAP_NAME}"

    echo "==> Copying snapshot out of ${LEADER_POD}..."
    kubectl -n vault cp "${LEADER_POD}:/tmp/${SNAP_NAME}" "/tmp/${SNAP_NAME}"
    kubectl -n vault exec "$LEADER_POD" -- rm -f "/tmp/${SNAP_NAME}"

    echo "==> Installing mc (MinIO client)..."
    curl -sf https://dl.min.io/client/mc/release/linux-amd64/mc -o /tmp/mc
    chmod +x /tmp/mc
    export MC_CONFIG_DIR=/tmp/.mc
    /tmp/mc alias set vps-minio "http://${VPS_MINIO_ENDPOINT}" \
      "${MINIO_ACCESS_KEY}" "${MINIO_SECRET_KEY}" >/dev/null

    echo "==> Uploading ${SNAP_NAME} to vps-minio/${BUCKET}..."
    /tmp/mc cp "/tmp/${SNAP_NAME}" "vps-minio/${BUCKET}/${SNAP_NAME}"
    rm -f "/tmp/${SNAP_NAME}"

    echo "==> Pruning snapshots older than 30 days..."
    /tmp/mc find "vps-minio/${BUCKET}" --older-than 30d --exec "/tmp/mc rm {}" || true

    echo "==> Done: ${SNAP_NAME}"
---
apiVersion: batch/v1
kind: CronJob
metadata:
  name: vault-raft-snapshot
  namespace: vault
spec:
  schedule: "0 */6 * * *"  # every 6 hours
  # Suspended 2026-08-17: the upload step targets vps-minio.netbird.internal,
  # which no pod in this cluster can currently reach - no node has an
  # egress route into the Netbird mesh (see commit history around this
  # date for the full investigation). The leader-detection fix above is
  # real and independent of this, so it's kept fixed rather than reverted.
  # Unsuspend once real pod-egress routing to the VPS exists.
  suspend: true
  concurrencyPolicy: Forbid
  successfulJobsHistoryLimit: 3
  failedJobsHistoryLimit: 3
  jobTemplate:
    spec:
      backoffLimit: 2
      template:
        spec:
          serviceAccountName: argocd-hook-sa  # already has kubectl exec rights in this namespace (see vault-init-job.yaml)
          restartPolicy: Never
          securityContext:
            runAsNonRoot: true
            runAsUser: 65534
            seccompProfile:
              type: RuntimeDefault
          containers:
            - name: raft-snapshot
              image: alpine/k8s:1.32.13
              command: ["/bin/bash", "/scripts/snapshot.sh"]
              securityContext:
                allowPrivilegeEscalation: false
                capabilities:
                  drop: ["ALL"]
              env:
                - name: MINIO_ACCESS_KEY
                  valueFrom:
                    secretKeyRef:
                      name: vps-minio-credentials
                      key: access-key
                - name: MINIO_SECRET_KEY
                  valueFrom:
                    secretKeyRef:
                      name: vps-minio-credentials
                      key: secret-key
              volumeMounts:
                - name: scripts
                  mountPath: /scripts
          volumes:
            - name: scripts
              configMap:
                name: vault-raft-snapshot-script
                defaultMode: 0755