diff options
| -rw-r--r-- | python/sandcrawler/ingest.py | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/python/sandcrawler/ingest.py b/python/sandcrawler/ingest.py index abcc156..eb8e256 100644 --- a/python/sandcrawler/ingest.py +++ b/python/sandcrawler/ingest.py @@ -26,6 +26,8 @@ from sandcrawler.db import SandcrawlerPostgrestClient from sandcrawler.xml import xml_reserialize +MAX_BODY_SIZE_BYTES = 128*1024*1024 + class IngestFileWorker(SandcrawlerWorker): """ High level flow is to look in history first, then go to live web if @@ -576,6 +578,10 @@ class IngestFileWorker(SandcrawlerWorker): result['status'] = 'null-body' return result + if len(resource.body) > MAX_BODY_SIZE_BYTES: + result['status'] = 'body-too-large' + return result + file_meta = gen_file_metadata(resource.body) try: file_meta, resource = fix_transfer_encoding(file_meta, resource) |