Details
-
Sub-task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
2.0.0
-
None
-
Reviewed
Description
Now we always use the default timeout configured for zk and meta.
I think it is reasonable to always use the same timeout when accessing zk and meta as the result will be shared by lots of threads. If we could do a successful fetching then the result will be in cache so it does not make sense to set the timeout to a very small value when fetching.
But I think it is also important to let the user request finish in time even if the user can only get a timeout exception. We should not block a user request longer than operation timeout.
So I think we could add a timeout parameter to the region locate method, and if the location can not be fetched in time, we will just finished the returned CompletableFuture with a timeout exception, but the actual fetching operation can still go on and use its own timeout config.