Closed
Description
Describe the bug
The size of the scroll_id token is proportional to the number of shards hit by the query.
We found info on scroll_id's generating large strings elastic/elasticsearch-py#971 . And looking over the Scroll code in rust, it always puts the scroll_id in to a GET request and in to the query params. So large queries will cause the Elasticsearch cluster's to report An HTTP line is larger than 4096 bytes.
.
To Reproduce
Steps to reproduce the behavior:
- Hit a lot of shards in a query
- Watch ES reject the second page of your scroll logic
Expected behavior
The Scroll code in rust should always use a POST so it can have long scroll_ids.
Screenshots
If applicable, add screenshots to help explain your problem.
Environment (please complete the following information):
- OS: [e.g. Windows 10 Pro]
- rustc version [e.g.
rustc --version
]
Additional context
Add any other context about the problem here.
Metadata
Metadata
Assignees
Labels
No labels