Closed dkijania closed 6 months ago
We received schema from third party partner. Will try to check if i can reproduce issue their schema
reproduced the same error without --repair-nonces
flag.
When running the replayer, I observed a few ledger hash mismatches from around January 2022. As I recall, there was some code change that was eventually reverted. After those few mismatches, all the ledger hashes should be as expected.
This wrong ledger corresponds to this orphaned block: https://minaexplorer.com/block/3NKm5VGDQXtekWf3erWfjcHciGzjAWB4u7WNE7gAHGYjvkFbF5xj and in our mainnet db dump, the user_commands and internal_commands for those 2 blocks are the same.
Yes, I recall that the mismatched hashes were associated with orphaned blocks.
If you run the replayer with --continue-on-error
, you'll see a few of these.
So this is caused by a minor bug that's already been fixed.
When running replayer around 146033 - 146752 slots, please use the --continue-on-error
flag and with --checkpoint-interval 1000
In the replayer log, you would expect to the following blocks at slot:
146033, 146078, 146102, 146164, 146328, 146399, 146438, 146489, 146631, 146752
have their ledger hash being "wrong".
Once a new checkpoint file is created around 147000 slot, then we can restart the replayer with this checkpoint file without --continue-on-error
flag anymore.
In the attempt of repair nonces on mainnet, we have found issue on both o1 and MF schemas. Reproduction path:
_repairnonces.sh
docker setup
Error