Details
-
Test
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
This is the aggregated jira to track all the integration test for standalone for 0.13.1
Test list:
- Leader failure should elect new leader
- Newly elected leader re-computes new JobModel and publishes it
- Processor with same id should not trigger a new JobModel & should stop itself after a timeout (new dies)
- Processor joining the group should timeout and stop, if there is no new JobModel available within a timeout interval
- Correct behavior when running #processors > #tasks (should not fail all processors, old ones stay)
- On barrier timeout we continue to work (test still needs to finish).
- Barrier completion without timeout
- Zk unused znodes are cleaned-up up to last 10 (multiple jobmodel versions, barriers etc)
- Handle session timeouts reliably
- ZK unavailability at start
- ZK failure (unavailable after start)
- Rolling upgrade of processors should not stop processing
- Processor leaving and joining the group within debounce time shouldn’t shuffle any of the tasks
- Works with high-level API
- Works with low-level (task API)
- Test Compatibility - supports stateless jobs with PassThroughJobCoordinator
- Test Compatibility - supports jobs with Brooklin Non-Loadbalanced System Consumer
- Test failure testing jobs run overnight (failure-testing.sh)
- Single StreamProcessor object
- Multiple StreamProcessor objects
- Adding another processor
- Adding >1 processors within the debounce time should only cause 1 JobModel recalculation
- Removing a processor
- Leader death and reelection
- Run with databus
- Single processor failure (internal failure, exception)
- StreamProcessor shuts down if Samza container does not shut down in time