-
Type:
Task
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: None
-
3
-
2023-06-27 Lord of the Sprints
If a thread caching data in the chunk cache takes too long to read from remote storage another thread waiting to access the same chunk times out and returns EAGAIN error up the call stack, which causes WiredTiger to crash.
We would like to change how this error is handled, because on high-latency storage devices we return EAGAIN too soon and crash too eagerly when there isn't in fact a problem.