Avoid concurrent indexing of the same document

mayan/apps/document_indexing/tasks.py

@@ -2,11 +2,13 @@ import logging
 
 from mayan.celery import app
 from documents.models import Document
+from lock_manager import Lock, LockError
 
 from .api import update_indexes, delete_indexes
 from .tools import do_rebuild_all_indexes
 
 logger = logging.getLogger(__name__)
+RETRY_DELAY = 20  # TODO: convert this into a config option
 
 
 @app.task(ignore_result=True)
@@ -15,12 +17,24 @@ def task_delete_indexes(document_id):
     delete_indexes(document)
 
 
-@app.task(ignore_result=True)
+@app.task(bind=True, ignore_result=True)
-def task_update_indexes(document_id):
+def task_update_indexes(self, document_id):
+    try:
+        lock = Lock.acquire_lock('document_indexing_task_update_index_document_%d' % document_id)
+    except LockError as exception:
+        # This document is being reindexed by another task, retry later
+        raise self.retry(exc=exception, countdown=RETRY_DELAY)
+    else:
+        try:
             document = Document.objects.get(pk=document_id)
+        except Document.DoesNotExist:
+            # Document was deleted before we could execute, abort about updating
+            pass
+        else:
             update_indexes(document)
 
 
 @app.task(ignore_result=True)
 def task_do_rebuild_all_indexes():
+    # TODO: Find a way to rebuild after all pending updates are finished
     do_rebuild_all_indexes()