Query timeout not working as expected

[bobbyi]

I am in the process of upgrading our cluster and client from ES 6.1.2 to ES 7.2 and have ran into some confusion regarding timeout settings with the Python client.

I have reduced down to a case where I create my client like this:
client = Elasticsearch([{'host': host, 'port': 9200}], timeout=60*60*5, max_retries=0, retry_on_timeout=False, sniff_on_start=True)

Note that I am specifying a timeout value of 5 hours (maybe too long for production, but this is for experimental purposes to diagnose what's happening here).

I then make a query like this:
client.search(index=index, body=query, size=0)
with no request_timeout setting (which should inherit the timeout value from the client), or alternately with an explicit request timeout of 5 hours:
client.search(index=index, body=query, size=0, request_timeout=60 * 60 * 5)

In either case, the query fails after ~45 seconds with a timeout exception:
ConnectionTimeout: ConnectionTimeout caused by - ReadTimeoutError(HTTPConnectionPool(host=u'10.110.182.171', port=9200): Read timed out. (read timeout=18000))

Note that the timeout value I specified is reflected in the error message.

If I've said the timeout should be 18000 seconds, why is it giving up after 45? How do I get my queries to wait as long as needed?

This is on Python 2.7.

[alexk307]

I believe this fixes the issue: https://github.com/elastic/elasticsearch-py/pull/1051/files

By removing the request_timeout parameter from the decorator, we no longer try to escape it which was causing an issue with urllib3. More details: #1049 (comment)

[sethmlarson]

I believe this issue is already resolved, will reopen if this is not the case.