Add final runtime and disk usage summary to mail reports

lib/common.sh

@@ -618,6 +618,7 @@ finish() {
       --hostname="$MAIL_HOSTNAME" --subject-prefix "$MAIL_SUBJECT_PREFIX" --run "$RUN_ID" --exit-code "$code" \
       --started "$RUN_STARTED_EPOCH" --phase "$BACKUP_PHASE" --upload-status "$UPLOAD_STATUS" \
       --config "$CONFIG_FILE" --workdir "$WORKDIR" --archive-dir "$ARCHIVE_DIR" \
+      --staging-root "$STAGING_ROOT" --restore-root "$RESTORE_ROOT" --log-dir "$LOG_DIR" \
       --compression "$COMPRESSOR" --remote "$RCLONE_REMOTE_BASE" --sendmail "$MAIL_SENDMAIL_BIN" \
       --tail-lines "$MAIL_INCLUDE_LOG_TAIL_LINES" --timeout "$MAIL_SEND_TIMEOUT" 2>&1 | tee -a "$LOG_FILE"
     rm -f -- "$report_sizes"

lib/report.py

@@ -16,6 +16,7 @@ import socket
 import subprocess
 import sys
 import time
+from usage import human, summary
 
 
 def allocated(path):
@@ -45,13 +46,6 @@ def snapshot(output, staging, archives):
     Path(output).write_text(json.dumps(data), encoding='utf-8')
 
 
-def human(value):
-    for unit in ('B', 'KiB', 'MiB', 'GiB', 'TiB'):
-        if value < 1024 or unit == 'TiB':
-            return f'{value:.1f} {unit}'
-        value /= 1024
-
-
 def deliver(args):
     fqdn = socket.getfqdn()
     host = args.hostname or fqdn
@@ -91,6 +85,8 @@ def deliver(args):
     else:
         body.append('Archive total: '+human(sum(sizes['archives'].values())))
     body.extend(['', f'Log-Auszug (letzte {args.tail_lines} Zeilen)', '-----------------', tail])
+    body.extend(summary(time.time() - args.started, args.workdir, args.archive_dir,
+                        args.staging_root, args.restore_root, args.log_dir))
     message = EmailMessage(policy=SMTP)
     # EmailMessage rejects CR/LF in headers and encodes non-ASCII safely.
     sender = args.sender or f'app-backup@{fqdn}'
@@ -117,6 +113,8 @@ def main():
         snap.add_argument(name)
     mail = commands.add_parser('deliver')
     mail.add_argument('--hostname', default='')
+    for name in ('staging-root', 'restore-root', 'log-dir'):
+        mail.add_argument('--'+name, required=True)
     for name in ('sizes', 'log', 'output', 'to', 'sender', 'subject-prefix', 'run', 'phase',
                  'upload-status', 'config', 'workdir', 'archive-dir', 'compression', 'remote', 'sendmail'):
         mail.add_argument('--'+name, required=True)

lib/restore_test.py

@@ -25,6 +25,7 @@ import tarfile
 import time
 
 import backup_set
+from usage import summary
 
 GIB = 1024 ** 3
 MARKER = 'app-backup-restore-test-v1\n'
@@ -530,6 +531,8 @@ exec mariadbd --no-defaults --user=root --datadir=/var/lib/mysql/db --socket=/va
                         f"Einträge {c.get('entries', '?')}; Tabellen {c.get('tables', '-')}; Phase {c.get('phase', '?')}")
         body.extend(['', 'Dateien entpackt; enthaltene DB-Dumps isoliert importiert und Tabellen geprüft.',
                      'Dies ist kein Start-/Funktionstest der produktiven Anwendungen.'])
+        body.extend(summary(time.time() - self.started, self.args.workdir, self.args.archive_dir,
+                            self.args.staging_root, self.args.restore_root, self.args.log_dir))
         message.set_content('\n'.join(body))
         raw = message.as_bytes()
         (self.root / 'report.eml').write_bytes(raw)
@@ -579,7 +582,8 @@ def io_bytes(raw):
 def main():
     parser = argparse.ArgumentParser()
     parser.add_argument('mode', choices=('run', 'check', 'cleanup', 'prepare'))
-    for key in ('workdir', 'prefix', 'remote', 'run', 'image', 'hostname', 'to', 'sender', 'sendmail', 'mail'):
+    for key in ('workdir', 'archive-dir', 'staging-root', 'restore-root', 'log-dir',
+                'prefix', 'remote', 'run', 'image', 'hostname', 'to', 'sender', 'sendmail', 'mail'):
         parser.add_argument('--' + key, required=True)
     for key in ('remote-check', 'config'):
         parser.add_argument('--' + key)

lib/usage.py

@@ -0,0 +1,72 @@
+"""Best-effort, bounded disk usage summary for backup and restore mail reports."""
+
+from pathlib import Path
+import os
+import subprocess
+
+
+def human(value):
+    for unit in ('B', 'KiB', 'MiB', 'GiB', 'TiB'):
+        if value < 1024 or unit == 'TiB':
+            return f'{value:.1f} {unit}'
+        value /= 1024
+
+
+def duration(seconds):
+    hours, remainder = divmod(max(0, int(seconds)), 3600)
+    minutes, seconds = divmod(remainder, 60)
+    return f'{hours}h {minutes:02d}m {seconds:02d}s'
+
+
+def allocated(path):
+    """Count allocated bytes without following symlinks or crossing filesystems."""
+    try:
+        result = subprocess.run(['du', '-sx', '-B1', '--', str(path)],
+                                capture_output=True, text=True, timeout=30, check=True)
+        return int(result.stdout.split()[0])
+    except (OSError, ValueError, IndexError, subprocess.SubprocessError):
+        return None
+
+
+def free_space(path):
+    try:
+        stat = os.statvfs(path)
+        return stat.f_bavail * stat.f_frsize
+    except OSError:
+        return None
+
+
+def summary(seconds, workdir, archives, staging, restore, logs):
+    """Measure the final local state, after temporary-file cleanup."""
+    lines = ['', 'Laufzeit und Speicher am Ende', '----------------------------',
+             f'Verarbeitung: {duration(seconds)} ({max(0, int(seconds))}s)']
+    paths = [('Arbeitsbereich gesamt', workdir), ('Backups (lokale Archive)', archives),
+             ('Backup-Staging', staging), ('Restore-Arbeitsbereich', restore),
+             ('Restore-Test', str(Path(workdir) / 'restore-test')), ('Protokolle und Berichte', logs)]
+    for label, name in paths:
+        if not name:
+            continue
+        path = Path(name)
+        amount = allocated(path) if path.exists() and not path.is_symlink() else None
+        lines.append(f'{label}: {human(amount) if amount is not None else "nicht verfügbar"} ({path})')
+    devices = set()
+    for label, name in [('Arbeitsbereich', workdir), ('Archive', archives),
+                        ('Staging', staging), ('Restore', restore), ('Restore-Test', str(Path(workdir) / 'restore-test')),
+                        ('Protokolle', logs)]:
+        if not name:
+            continue
+        try:
+            path = Path(name)
+            if path.is_symlink():
+                continue
+            device = path.stat().st_dev
+        except OSError:
+            continue
+        if device in devices:
+            continue
+        devices.add(device)
+        amount = free_space(path)
+        lines.append(f'Frei auf Datenträger {label}: '
+                     f'{human(amount) if amount is not None else "nicht verfügbar"} ({path})')
+    lines.append('Teilbereiche sind im Arbeitsbereich enthalten; getrennte Datenträger werden separat gemessen.')
+    return lines