-
Type:
Task
-
Resolution: Fixed
-
Priority:
Minor - P4
-
Affects Version/s: None
-
Component/s: Replication
-
Fully Compatible
-
v4.4
-
Repl 2020-04-20, Repl 2020-05-04
We should investigate the slowest parts of the RollbackTest fixture to identify and remove any areas of wasted downtime. Wasted downtime would be any time spent not doing useful, required work. For example, waiting for data to propagate over the network or waiting for an arbitrary timeout to fire in the server e.g. a heartbeat interval or election timeout.
In many of our tests that use the rollback test fixture, the duration of the rollback itself is often only a few hundred milliseconds. In theory, our test fixtures should be able to run a rollback test sequence in a time not much longer than it takes to complete a rollback, setting aside required setup and teardown time. We should investigate to see if this is feasible with our current testing infrastructure.
- is duplicated by
-
-
- Closed
-
- is related to
-
SERVER-47702 Tests should not try to resume index builds via failpoint while a node may be in ROLLBACK
-
- Closed
-
-
-
- Closed
-
-
SERVER-47814 Tests that pass a custom ReplSetTest to RollbackTest should enable the 'setSmallOplogGetMoreMaxTimeMS' failpoint
-
- Closed
-