Closed edward-swirldslabs closed 1 week ago
PCES snapshots are skipped for signature generation.
Perhaps we can add a skip to validation under the same condition.
if (reservedState.isComplete() || reservedState.isPcesRound()) {
// when state is complete or signature generation has been skipped for PCES generated state snapshots, nothing to do
return;
}
PCES snapshots are skipped for signature generation.
Perhaps we can add a skip to validation under the same condition.
if (reservedState.isComplete() || reservedState.isPcesRound()) { // when state is complete or signature generation has been skipped for PCES round states, nothing to do return; }
I think this is a fine solution, to get rid of the nuisance log. No problems come to mind, in any case.
It doesn't fix the problem of having an unsigned snapshot on disk (though I'm not sure how much of a problem that even is...). But at least we can stop spending time debugging these tests
Description
JRS Test Results: http://35.247.76.217:8095/swirlds-automation/release/0.53/4N_2C/Ubuntu2204_Update/20240825-111009-GCP-Daily-Update-Ubuntu2204-4N-2C/Crypto-Update-Jar-1.5k-25m/
Every node wrote a state to disk on round 4217 at 46.6 seconds with 0 of 501 stake accounted for. The weight distribution on nodes is as follows:
It could be the case there were no signatures on the state, or 1 signature from node id 3.
Given how rare this event is, it regularly happens that node 0 provides a signature to the state. In this case node 0, 1, and 2 failed to have signatures present.
signing weight is initialized to 0 here:
When the signatures are set
When invalid signatures are pruned
While signatures on state are best effort and not guaranteed, we should investigate further to see if we can find an explanation more satisfying than "it randomly happens."
The platform loaded a state from round 4216
PCES Replay reached 4,219
The problem happened on round 4217
PCES Replay created the snapshot. Either we did not create the signatures after upgrade, or the signatures were never produced prior to upgrade. Or they were dropped.
Steps to reproduce
Follow the JRS test and pray to RNGesus it happens with a debugger attached.
Additional context
Debug Address Book Before Update:
Debug Address Book After Update:
usedAddressBook_vHederaSoftwareVersion{hapiVersion=v0.53.0, servicesVersion=v0.80.0}_2024-08-25-11-33-46_node_3.debug.txt
Hedera network
other
Version
v0.53.x
Operating system
Linux