feat: retry urls with connection errors

cms/djangoapps/contentstore/tasks.py

@@ -1120,8 +1120,10 @@ def validate_user():
             return
 
     def get_urls(content):
-        """Returns all urls foundafter href and src in content.
-        Excludes urls that are only '#'."""
+        """
+        Returns all urls found after href and src in content.
+        Excludes urls that are only '#'.
+        """
         regex = r'\s+(?:href|src)=["\'](?!#)([^"\']*)["\']'
         url_list = re.findall(regex, content)
         return url_list
@@ -1149,9 +1151,33 @@ def convert_to_standard_url(url, course_key):
         else:
             return url
 
+    def scan_course_for_links(course_key):
+        """
+        Returns a list of all urls in a course.
+        Returns: [ [block_id1, url1], [block_id2, url2], ... ]
+        """
+        verticals = modulestore().get_items(course_key, qualifiers={'category': 'vertical'}, revision=ModuleStoreEnum.RevisionOption.published_only)
+        blocks = []
+        urls_to_validate = []
+
+        for vertical in verticals:
+            blocks.extend(vertical.get_children())
+
+        for block in blocks:
+            block_id = str(block.usage_key)
+            block_info = get_block_info(block)
+            block_data = block_info['data']
+
+            url_list = get_urls(block_data)
+            urls_to_validate += [[block_id, url] for url in url_list]
+
+        return urls_to_validate
+
     async def validate_url_access(session, url_data, course_key):
-        """Returns status of a url request.
-        url_list is [id, url]"""
+        """
+        Returns the status of a url request
+        Returns: {block_id1, url1, status}
+        """
         block_id, url = url_data
         result = {'block_id': block_id, 'url': url}
         standardized_url = convert_to_standard_url(url, course_key)
@@ -1160,12 +1186,14 @@ async def validate_url_access(session, url_data, course_key):
                 result.update({'status': response.status})
         except Exception as e:
             result.update({'status': None})
-            print('[Validate url error]', type(e), e, url)
+            LOGGER.debug(f'[Link Check] Request error when validating {url}: {str(e)}')
         return result
 
     async def validate_urls_access_in_batches(url_list, course_key, batch_size=100):
-        """Returns the statuses of a list of url requests.
-        url_list is [block_id, url]"""
+        """
+        Returns the statuses of a list of url requests.
+        Returns: [ {block_id1, url1, status}, {block_id2, url2, status}, ... ]
+        """
         responses = []
         url_count = len(url_list)
 
@@ -1175,67 +1203,72 @@ async def validate_urls_access_in_batches(url_list, course_key, batch_size=100):
                 tasks = [validate_url_access(session, url_data, course_key) for url_data in batch]
                 batch_results = await asyncio.gather(*tasks)
                 responses.extend(batch_results)
-                print(f'batch {i // batch_size+1} of {url_count // batch_size + 1}')
+                LOGGER.debug(f'[Link Check] request batch {i // batch_size+1} of {url_count // batch_size + 1}')
 
         return responses
 
-    def scan_course_for_links(course_key):
+    def filter_by_status(results):
         """
-        Returns a list of links that are broken or locked.
-          [block_id, link, is_locked]
+        Filter results by status.
+            200: OK. No need to do more
+            403: Forbidden. Record as locked link.
+            None: Error. Retry up to 3 times.
+            Other: Failure. Record as broken link.
+        Returns:
+            filtered_results: [ [block_id1, url1, is_locked], ... ]
+            retry_list: [ [block_id1, url1], ... ]
         """
-        verticals = modulestore().get_items(course_key, qualifiers={'category': 'vertical'}, revision=ModuleStoreEnum.RevisionOption.published_only)
-        blocks = []
-        links_to_validate = []
-
-        for vertical in verticals:
-            blocks.extend(vertical.get_children())
-
-        for block in blocks:
-            block_id = str(block.usage_key)
-            block_info = get_block_info(block)
-            block_data = block_info['data']
-
-            url_list = get_urls(block_data)
-            links_to_validate += [[block_id, url] for url in url_list]
-
-        return links_to_validate
+        filtered_results = []
+        retry_list = []
+        for result in results:
+            if result['status'] == None:
+                retry_list.append([result['block_id'], result['url']])
+            elif result['status'] == 200:
+                continue
+            elif result['status'] == 403 and is_studio_url(result['url']):
+                filtered_results.append([result['block_id'], result['url'], True])
+            else:
+                filtered_results.append([result['block_id'], result['url'], False])
+
+        return filtered_results, retry_list
 
     user = validate_user()
 
     self.status.set_state('Scanning')
     course_key = CourseKey.from_string(course_key_string)
-    links_list = scan_course_for_links(course_key)
-    results = asyncio.run(validate_urls_access_in_batches(links_list, course_key, batch_size=100))
-
-    final_results = []
-    for result in results:
-        if result['status'] == None: # Request error
-            print('retry') # TODO retry
-        if result['status'] == 200: # OK
-            print('remove from list') # TODO remove
-        elif result['status'] == 403 and is_studio_url(result['url']):
-            final_results.append([result['block_id'], result['url'], True])
-        else:
-            final_results.append([result['block_id'], result['url'], False])
+    url_list = scan_course_for_links(course_key)
+    validated_url_list = asyncio.run(validate_urls_access_in_batches(url_list, course_key, batch_size=100))
+    broken_or_locked_urls, retry_list = filter_by_status(validated_url_list)
+
+    # Retry urls that failed due to connection error
+    retry_count = 3
+    for i in range(0, retry_count):
+        if retry_list:
+            LOGGER.debug(f'[Link Check] retry attempt #{i+1}')
+            retry_validated_url_list = asyncio.run(validate_urls_access_in_batches(retry_list, course_key, batch_size=100))
+            retry_results, retry_list = filter_by_status(retry_validated_url_list)
+            broken_or_locked_urls.extend(retry_results)
+
+    if retry_list:
+        LOGGER.debug(f'[Link Check] {len(retry_list)} requests failed due to connection error')
 
     try:
         self.status.increment_completed_steps()
 
         file_name = str(course_key)
-        links_file = NamedTemporaryFile(prefix=file_name + '.', suffix='.json')
-        LOGGER.debug('json file being generated at %s', links_file.name)
+        broken_links_file = NamedTemporaryFile(prefix=file_name + '.', suffix='.json')
+        LOGGER.debug(f'[Link Check] json file being generated at {broken_links_file.name}')
 
-        with open(links_file.name, 'w') as file:
-            json.dump(final_results, file, indent=4)
+        with open(broken_links_file.name, 'w') as file:
+            json.dump(broken_or_locked_urls, file, indent=4)
 
         artifact = UserTaskArtifact(status=self.status, name='BrokenLinks')
-        artifact.file.save(name=os.path.basename(links_file.name), content=File(links_file))
+        artifact.file.save(name=os.path.basename(broken_links_file.name), content=File(broken_links_file))
         artifact.save()
 
     # catch all exceptions so we can record useful error messages
-    except Exception as exception:  # pylint: disable=broad-except
+    except Exception as e:  # pylint: disable=broad-except
         LOGGER.exception('Error checking links for course %s', course_key, exc_info=True)
         if self.status.state != UserTaskStatus.FAILED:
-            self.status.fail({'raw_error_msg': str(exception)})
+            self.status.fail({'raw_error_msg': str(e)})
         return