Feature: Prometheus metrics for `rustic forget` and `rustic forget --prune`

[zoechi]

I just learned that metrics are only emitted for rustic backup but not for rustic forget #1505 (reply in thread)

Summary

rustic backup emits a rich set of Prometheus metrics (timing, data volume, blob counts, file/dir change stats). rustic forget and rustic forget --prune currently emit no metrics, creating an observability blind spot: you cannot alert on forgotten snapshots, track storage reclamation, or measure prune operation duration.

This issue proposes a set of metrics for both subcommands, modeled on the existing rustic_backup_* namespace.

---

Existing backup metrics — mapping to forget

✅ Direct equivalents (same concept, renamed)

Backup metric	Proposed forget equivalent	Notes
rustic_backup_time	rustic_forget_time	Unix timestamp when operation started
rustic_backup_backup_start	rustic_forget_forget_start	Unix timestamp of forget phase start
rustic_backup_backup_end	rustic_forget_forget_end	Unix timestamp of forget phase end
rustic_backup_backup_duration	rustic_forget_forget_duration	Duration of forget phase (seconds)
rustic_backup_total_duration	rustic_forget_total_duration	Total duration incl. prune phase (seconds)

For --prune, add a separate timing pair for the prune phase:

Proposed metric	Notes
rustic_forget_prune_start	Unix timestamp of prune phase start
rustic_forget_prune_end	Unix timestamp of prune phase end
rustic_forget_prune_duration	Duration of prune phase only (seconds)

---

↔️ Opposite-direction equivalents (prune only)

Backup adds data; prune removes it. These mirror the data_added family but in reverse. Only emitted when --prune is active.

Backup metric	Proposed forget/prune equivalent	Notes
rustic_backup_data_added	rustic_forget_data_removed	Total bytes freed (uncompressed)
rustic_backup_data_added_files	rustic_forget_data_removed_files	File bytes freed (uncompressed)
rustic_backup_data_added_trees	rustic_forget_data_removed_trees	Tree/dir bytes freed (uncompressed)
rustic_backup_data_added_packed	rustic_forget_data_removed_packed	Storage bytes freed (after compression)
rustic_backup_data_added_files_packed	rustic_forget_data_removed_files_packed	File storage freed (compressed)
rustic_backup_data_added_trees_packed	rustic_forget_data_removed_trees_packed	Tree storage freed (compressed)
rustic_backup_data_blobs	rustic_forget_data_blobs_removed	Count of data blobs deleted
rustic_backup_tree_blobs	rustic_forget_tree_blobs_removed	Count of tree blobs deleted

---

❌ Not applicable for forget

These backup metrics relate to filesystem scanning, which forget does not perform — no equivalents needed:

• rustic_backup_files_new / _changed / _unmodified
• rustic_backup_dirs_new / _changed / _unmodified
• rustic_backup_total_files_processed
• rustic_backup_total_dirs_processed
• rustic_backup_total_bytes_processed
• rustic_backup_total_dirsize_processed

---

New forget-specific metrics

Snapshot counts (core purpose of forget)

Metric	Notes
rustic_forget_snapshots_total	Total snapshots in repo before forget
rustic_forget_snapshots_removed	Snapshots removed this run
rustic_forget_snapshots_kept	Snapshots kept this run

Retention policy breakdown

Knowing why snapshots were kept helps validate policy configuration. Either as separate metrics or a single metric with a reason label:

rustic_forget_snapshots_kept{reason="last"}
rustic_forget_snapshots_kept{reason="hourly"}
rustic_forget_snapshots_kept{reason="daily"}
rustic_forget_snapshots_kept{reason="weekly"}
rustic_forget_snapshots_kept{reason="monthly"}
rustic_forget_snapshots_kept{reason="yearly"}
rustic_forget_snapshots_kept{reason="tag"}
rustic_forget_snapshots_kept{reason="within"}

A single labeled metric is preferable to 8 separate metrics — easier to query with sum by (reason) and fewer time series.

Pack file metrics (prune only)

Metric	Notes
rustic_forget_packs_removed	Pack files fully deleted (all blobs unused)
rustic_forget_packs_rewritten	Pack files rewritten (partial blob removal)
rustic_forget_packs_kept	Pack files left untouched

---

Full proposed metric list

rustic forget (without --prune)

rustic_forget_time
rustic_forget_forget_start
rustic_forget_forget_end
rustic_forget_forget_duration
rustic_forget_total_duration
rustic_forget_snapshots_total
rustic_forget_snapshots_removed
rustic_forget_snapshots_kept{reason="..."}

Additional metrics for rustic forget --prune

rustic_forget_prune_start
rustic_forget_prune_end
rustic_forget_prune_duration
rustic_forget_data_removed
rustic_forget_data_removed_files
rustic_forget_data_removed_trees
rustic_forget_data_removed_packed
rustic_forget_data_removed_files_packed
rustic_forget_data_removed_trees_packed
rustic_forget_data_blobs_removed
rustic_forget_tree_blobs_removed
rustic_forget_packs_removed
rustic_forget_packs_rewritten
rustic_forget_packs_kept

---

Alerting use cases enabled

With these metrics, the following Prometheus alerts become possible:

• Forget not running: time() - rustic_forget_time > 86400
• Prune taking too long: rustic_forget_prune_duration > 3600
• Snapshot count growing unbounded: rustic_forget_snapshots_total > threshold
• Policy misconfiguration (no snapshots being removed): rustic_forget_snapshots_removed == 0 sustained over time
• Storage not being freed despite removals: rustic_forget_data_removed_packed == 0 when rustic_forget_snapshots_removed > 0