I am using the python library to do a scan/scroll of an index with ~12 million documents, do some work on them, and then bulk index them into a new index.
I run this script nightly on a cron job and sometimes (approximately 10% of the time), I get the following error: "ElasticsearchIllegalStateException[No matching token for number_type [BIG_INTEGER]]" during the scan.
There is no corresponding error in the server logs, and none of the documents being scanned have mapped fields that are the "long" (which corresponds to BIG_INTEGER I believe) type.
I have investigated the documents that are failing during the scan by ID, and retrieving them by ID directly or by query on the _id field yields no issues.
I have also played around with the scroll time and fetch size parameters to no avail.
I am completely out of ideas and would be very grateful if anyone had any insight, or could point me in the right direction. Thanks!