fix cleanup job to delete orphaned seed files after org has been deleted (#2920)

- in seed cleanup, catch if org doesn't exist and attempt to delete file
directly
- add delete_file_from_default_storage() to allow deleting left-over
seed files, assuming they are stored in default storage
- side effect of #2918, which will be fixed in #2919
- no need for migration, as will be picked up existing cleanup seed
files job

Author: ikreymer

backend/btrixcloud/file_uploads.py

@@ -198,9 +198,12 @@ async def upload_user_file_stream(
 
         file_id = uuid4()
 
-        new_filename = f"{upload_type}-{str(file_id)}{extension}"
+        new_filename = f"{str(file_id)}{extension}"
 
-        prefix = org.storage.get_storage_extra_path(str(org.id)) + f"{upload_type}s/"
+        prefix = (
+            org.storage.get_storage_extra_path(str(org.id))
+            + f"{upload_type}s/{upload_type}"
+        )
 
         file_prep = UserFilePreparer(
             prefix,
@@ -383,6 +386,22 @@ async def cleanup_unused_seed_files(self):
                 org = await self.org_ops.get_org_by_id(file_dict["oid"])
                 await self.delete_seed_file(file_id, org)
                 print(f"Deleted unused seed file {file_id}", flush=True)
+
+            except HTTPException as e:
+                # handle case where org is deleted by seed file still exists
+                if e.detail == "invalid_org_id":
+                    try:
+                        await self.storage_ops.delete_file_from_default_storage(
+                            file_dict["filename"]
+                        )
+                        await self.files.delete_one({"_id": file_id})
+                    # pylint: disable=broad-exception-caught
+                    except Exception as err:
+                        print(
+                            f"Error deleting orphaned seed file without org {file_id}: {err}",
+                            flush=True,
+                        )
+
             # pylint: disable=broad-exception-caught
             except Exception as err:
                 print(f"Error deleting unused seed file {file_id}: {err}", flush=True)

backend/btrixcloud/storages.py

@@ -73,7 +73,7 @@
 
 # ============================================================================
 # pylint: disable=broad-except,raise-missing-from,too-many-instance-attributes
-# pylint: disable=too-many-public-methods
+# pylint: disable=too-many-public-methods, too-many-lines
 class StorageOps:
     """All storage handling, download/upload operations"""
 
@@ -638,17 +638,32 @@ async def _delete_file(
         self, org: Organization, filename: str, storage: StorageRef
     ) -> bool:
         """delete specified file from storage"""
-        status_code = None
-
         s3storage = self.get_org_storage_by_ref(org, storage)
 
+        return await self._delete_file_from_storage(s3storage, filename)
+
+    async def _delete_file_from_storage(self, s3storage: S3Storage, filename: str):
+        """delete file from specified storage"""
+        status_code = None
+
         async with self.get_s3_client(s3storage) as (client, bucket, key):
             key += filename
             response = await client.delete_object(Bucket=bucket, Key=key)
             status_code = response["ResponseMetadata"]["HTTPStatusCode"]
 
         return status_code == 204
 
+    async def delete_file_from_default_storage(self, filename: str):
+        """delete file from default primary storage, if it exists"""
+        if not self.default_primary:
+            return False
+
+        s3storage = self.default_storages.get(self.default_primary.name)
+        if not s3storage:
+            return False
+
+        return await self._delete_file_from_storage(s3storage, filename)
+
     async def sync_stream_wacz_pages(
         self, wacz_files: List[CrawlFileOut], num_retries=5
     ) -> Iterator[Dict[Any, Any]]:

scripts/build-backend.sh

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
 CURR=$(dirname "${BASH_SOURCE[0]}")
 
-docker build -t ${REGISTRY}webrecorder/browsertrix-backend:latest $CURR/../backend/
+docker build --load -t ${REGISTRY}webrecorder/browsertrix-backend:latest $CURR/../backend/
 
 if [ -n "$REGISTRY" ]; then
     docker push ${REGISTRY}webrecorder/browsertrix-backend