Closed jpraynaud closed 2 months ago
After investigation, we were able to understand and reproduce the problem:
It's caused by the chunk import system with a sequence of blocks containing no transactions. The chunk import divides the total number of blocks to be imported by 1500. In each iteration, the lower bound (starting block) remains the same, but the range of blocks to be imported grows larger. This behavior is only reproducible on a network like Sanchonet, which contains very few transactions.
On testing-preview, the import time is about 45 minutes with 4,793,471 transactions (included in 1,234,225 blocks in a range of 2,313,659 blocks). On testing-sanchonet, the import time is about 25-30 hours with 8,932 transactions (included 4,897 blocks in a range of 1,742,609 blocks)
Why
We have noticed that the signature of the Cardano transactions is not working properly on the
testing-sanchonet
network.The error message received is:
After investigation, it appears that the signer node does not always sign the same state than the aggregator (which leads to having the signature rejected by the aggregator).
After resetting the Cardano transactions databases of the signer and the aggregator on the
testing-sanchonet
we have noticed that the transaction importer was stuck and kept trying to import the same blocks:After each round of the importer, the
cardano_tx
table is left unchanged:Also, we have noticed that a test SPO (which is not run by the core team) with the same version of the signer
0.2.167
is able to sign properly the transactions.What
Investigate and fix the problem(s) which prevent the signature of the Cardano transactions.
How
Fix the problem(s) and make thetesting-sanchonet
network sign the Cardano transactions