Steps to Reproduce:
1. Set up the pool.
2. Run load test with load near to production.
3. Stop several nodes (Node13 and Node20 for example).
4. Stop the primary to initiate View Change.
5. Start stopped nodes (one by one, including primary).
6. Reduce load rate of script.
7. Wait for the end of catch up on nodes which were stopped.
8. Try to write several txns.
Pool can write, but nodes which were stopped don't order txns. The last message with "ordered batch request" was written to logs on view 0:
2019-01-17 12:58:04,538|INFO|replica.py|Node13:6 ordered batch request, view no 0, ppSeqNo 92, ledger 1001, state root None, txn root None, requests ordered 2, discarded 0
Nodes should order after View Change and Catch Up.
Logs and metrics: s3://qanodelogs/indy-1876/not_ordering_after_VC_and_catchup_17_01_2019
To get logs, run following command on log processor machine:
aws s3 cp --recursive s3://qanodelogs/indy-1876/not_ordering_after_VC_and_catchup_17_01_2019/ /home/ev/logs/qanodelogs/indy-1876/not_ordering_after_VC_and_catchup_17_01_2019/