-
Notifications
You must be signed in to change notification settings - Fork 8.5k
Description
Kibana version:
7.6, 7.6.1, 8.0
Elasticsearch version:
Server OS version:
Browser version:
Browser OS version:
Original install method (e.g. download page, yum, from source, etc.):
Describe the bug:
This bug is a combination of several issues:
-
This pr accidentally removed the timeout from the autocomplete request making the autocomplete requests run without a timeout (combination of dropping the
mssuffix from timeout and fetching the config from the wrong place, resulting in an undefined timeout being sent to ES. -
This pr introduced query cancellation, but the query wasn't actually canceled on the server until 7.7 Abort cancelled search requests to Elasticsearch #56788
Together, these two issues cause ES to get to 100% CPU in a production environment, making it almost unusable unless KQL autocomplete is disabled.
Steps to reproduce:
- Run kibana in verbose mode
elasticsearch.logQueries: true
logging.verbose: true
- Type a KQL query into the query bar.
- Find the request sent to ES in the console. it should have had
timeoutandterminate_after, but it doesn't.
Expected behavior:
All requests should have timeout and terminate_after
Screenshots (if relevant):
Errors in browser console (if relevant):
Provide logs and/or server output (if relevant):
Any additional context: