Because checkpoint rollback is triggered by intended instance failure at a given component when the second checkpoint control tuple arrives. The failed instances will be restarted then and an additional checkpoint will be made afterwards which will result in unexpected instance state info. What happens to the instances which can triggered checkpoint rollback: 1. instance started 2. successfully make checkpoint 1 3. throw exception to trigger instance failure when making checkpoint 2 4. restart this instance and rollback to checkpoint 1 5. successfully make checkpoint 3 6. throw exception to trigger instance failure when making checkpoint 4 ...... the expected results are the instance states in checkpoint 1, so only those info at the http server is expected. At the instance level, checkpoints will be made periodically and cannot be stopped or otherwise controlled.
[ Full content available at: https://github.com/apache/incubator-heron/pull/3004 ] This message was relayed via gitbox.apache.org for [email protected]
