Merge branch '378-add-health-check-for-dbbackup' into 'master'

Resolve "Add health check for dbbackup"

Closes #378

See merge request !495

aleksis/core/apps.py

@@ -53,9 +53,17 @@ class CoreConfig(AppConfig):
 
         self._load_data_checks()
 
-        from .health_checks import DataChecksHealthCheckBackend
+        from .health_checks import (
+            BackupJobHealthCheck,
+            DataChecksHealthCheckBackend,
+            DbBackupAgeHealthCheck,
+            MediaBackupAgeHealthCheck,
+        )
 
         plugin_dir.register(DataChecksHealthCheckBackend)
+        plugin_dir.register(DbBackupAgeHealthCheck)
+        plugin_dir.register(MediaBackupAgeHealthCheck)
+        plugin_dir.register(BackupJobHealthCheck)
 
     @classmethod
     def _load_data_checks(cls):

aleksis/core/health_checks.py

+from datetime import datetime
+
+from django.conf import settings
 from django.utils.translation import gettext as _
 
+from dbbackup import utils as dbbackup_utils
+from dbbackup.storage import get_storage
+from django_celery_results.models import TaskResult
 from health_check.backends import BaseHealthCheckBackend
 
 from aleksis.core.models import DataCheckResult
@@ -16,3 +22,54 @@ class DataChecksHealthCheckBackend(BaseHealthCheckBackend):
 
     def identifier(self):
         return self.__class__.__name__
+
+
+class BaseBackupHealthCheck(BaseHealthCheckBackend):
+    """Common base class for backup age checks."""
+
+    critical_service = False
+    content_type = None
+    configured_seconds = None
+
+    def check_status(self):
+        storage = get_storage()
+        backups = storage.list_backups(content_type=self.content_type)
+        if backups:
+            last_backup = backups[-1]
+            last_backup_time = dbbackup_utils.filename_to_date(last_backup)
+            time_gone_since_backup = last_backup_time - datetime.now()
+
+            # Check if backup is older than configured time
+            if time_gone_since_backup.seconds > self.configured_seconds:
+                self.add_error(_(f"Last backup {time_gone_since_backup}!"))
+        else:
+            self.add_error(_("No backup found!"))
+
+
+class DbBackupAgeHealthCheck(BaseBackupHealthCheck):
+    """Checks if last backup file is less than configured seconds ago."""
+
+    content_type = "db"
+    configured_seconds = settings.DBBACKUP_CHECK_SECONDS
+
+
+class MediaBackupAgeHealthCheck(BaseBackupHealthCheck):
+    """Checks if last backup file is less than configured seconds ago."""
+
+    content_type = "media"
+    configured_seconds = settings.MEDIABACKUP_CHECK_SECONDS
+
+
+class BackupJobHealthCheck(BaseHealthCheckBackend):
+    """Checks if last backup file is less than configured seconds ago."""
+
+    critical_service = False
+
+    def check_status(self):
+        task = TaskResult.objects.filter(task_name="aleksis.core.tasks.backup_data").last()
+
+        # Check if state is success
+        if not task:
+            self.add_error(_("No backup result found!"))
+        elif task and task.status != "SUCCESS":
+            self.add_error(_(f"{task.status} - {task.result}"))

aleksis/core/settings.py

@@ -730,4 +730,7 @@ HEALTH_CHECK = {
     "MEMORY_MIN": _settings.get("health.memory_min_mb", 500),
 }
 
+DBBACKUP_CHECK_SECONDS = _settings.get("backup.database.check_seconds", 7200)
+MEDIABACKUP_CHECK_SECONDS = _settings.get("backup.media.check_seconds", 7200)
+
 PROMETHEUS_EXPORT_MIGRATIONS = False