Details
-
Bug
-
Resolution: Won't Fix
-
Critical
-
6.0.4
-
Untriaged
-
1
-
No
Description
Looks like the query timeout was never implemented or handled correctly with index aliases. The current implementation fails to cancel queries upon the expiry of the given timeout value in query when fired against an index alias.
With index aliases, we are not using the correct api which takes a cancellable context. Hence the user given timeout value is never taken further to the remote/local partitions and those will assume a timeout value of zero which means “no timeout”.
This prevents any timeout/ expiry based cancellation of queries to leave the possibility of orphan/stale queries handles.
Fixing this certainly can help to avoid the query handle leaks as well as put a threshold on the server resource usages by strictly respecting any the query latency SLAs given as query timeouts.
Attachments
Issue Links
- is a backport of
-
MB-39949 FTS - Incomplete Query timeout handling for index aliases
- Closed