Affects Version/s: None
Fix Version/s: 1.6
Sprint:EV 18.14 Monitoring/Stability
As was found in
INDY-1404, if a lot of catchup-related messages (LEDGER_STATUS, CONSISTENCY_PROOFS) are lost, the node doesn't try to start catchup again.
The messages can be lost due to high load (as in the test), and limited size of node-to-node's message queue.
The recent fixes regarding catchup and fixes done in the scope of
INDY-1404 made the problem more visible, since we require that the previous round of catchup is finished before starting the next round.
- The following tests need to be written:
- Skip sending of Message Responses to LEDGER STATUS request once for domain ledger to a node doing catchup; make sure that the node eventually finishes catchup
- Skip sending of CONSISTENCY_PROOFs for domain ledger once to a node doing catchup; make sure that the node eventually finishes catchup
- A Node must finish catchup regardless of the load in the pool
- Catchup can be finished with some delay (in a timeout),