fleet-manager: add backup.sh for the durable-state volumes

Snapshot fleet-data (inventory + admins.json + enroll.key — the admin-lockout
risk) and influx-data to timestamped tarballs via a throwaway alpine container
(read-only, live-safe), with BACKUP_KEEP retention. Documented in DEPLOY.md
Backups with a restore pointer.

serving/fleet-manager/DEPLOY.md

@@ -222,6 +222,19 @@ and the `influx-data` volume. Losing `admins.json` locks every admin out — you
 re-bootstrap with `register-admin`. Losing `enroll.key` invalidates any
 outstanding (unused) enrol links.
 
+`./backup.sh [dest-dir]` tars both volumes (read-only, safe to run live) into
+timestamped `.tar.gz` files (default `./backups`, keeps the newest
+`BACKUP_KEEP=14`). Schedule it with a systemd timer / cron and copy the output
+offsite:
+
+```bash
+./backup.sh /srv/backups/fleet-manager
+# restore (manager stopped) — see the header of backup.sh
+```
+
+If you run compose under a non-default project name, pass
+`COMPOSE_PROJECT_NAME=<name>` so it finds the `<name>_fleet-data` volume.
+
 ## Troubleshooting
 
 - **`permission denied` writing the inventory on approve** — the runtime image is

serving/fleet-manager/backup.sh

@@ -0,0 +1,56 @@
+#!/usr/bin/env bash
+# backup.sh — snapshot the fleet-manager durable state to timestamped tarballs.
+#
+# Two docker named volumes hold everything that isn't reproducible from the image:
+#   - fleet-data  : inventory.json + admins.json + enroll.key  (CRITICAL — losing
+#                   admins.json/enroll.key locks every admin out of the UI)
+#   - influx-data : the metrics time-series (history; regenerates going forward)
+#
+# Runs a throwaway alpine container to tar each volume read-only into $DEST. Safe
+# to run live. Schedule it (systemd timer / cron); copy $DEST offsite.
+#
+# Usage:
+#   ./backup.sh [dest-dir]                 # default dest: ./backups
+#   COMPOSE_PROJECT_NAME=foo ./backup.sh   # if you run compose under another project
+#
+# Restore (manager stopped):
+#   docker run --rm -v <project>_fleet-data:/dest -v "$PWD/backups":/src alpine \
+#     sh -c 'rm -rf /dest/* && tar xzf /src/fleet-data-YYYYmmdd-HHMMSS.tar.gz -C /dest'
+set -euo pipefail
+
+DEST=${1:-./backups}
+PROJECT=${COMPOSE_PROJECT_NAME:-fleet-manager}   # compose derives volume names as <project>_<key>
+STAMP=$(date -u +%Y%m%d-%H%M%S)
+VOLUMES=${BACKUP_VOLUMES:-"fleet-data influx-data"}
+
+mkdir -p "$DEST"
+DEST_ABS=$(cd "$DEST" && pwd)
+
+for key in $VOLUMES; do
+    vol="${PROJECT}_${key}"
+    if ! docker volume inspect "$vol" >/dev/null 2>&1; then
+        echo "backup: SKIP $vol (no such volume — wrong COMPOSE_PROJECT_NAME?)"
+        continue
+    fi
+    out="${key}-${STAMP}.tar.gz"
+    echo "backup: $vol -> $DEST_ABS/$out"
+    docker run --rm \
+        -v "${vol}:/src:ro" \
+        -v "${DEST_ABS}:/backup" \
+        alpine tar czf "/backup/${out}" -C /src .
+done
+
+# Retention: keep the newest $KEEP of each volume's tarballs (0 = keep all).
+KEEP=${BACKUP_KEEP:-14}
+if [ "$KEEP" -gt 0 ]; then
+    for key in $VOLUMES; do
+        # ls -t for mtime order; our filenames are safe (key + UTC timestamp).
+        # shellcheck disable=SC2012
+        ls -1t "$DEST_ABS/${key}-"*.tar.gz 2>/dev/null | tail -n +"$((KEEP + 1))" | while read -r old; do
+            echo "backup: prune $old"
+            rm -f "$old"
+        done
+    done
+fi
+
+echo "backup: done ($STAMP) — copy $DEST_ABS offsite."