Node crashes when rolling back

[abailly]

What revision are you using?

b7d9fce

What operating system are you using?

Linux

What architecture are you using?

x86_64/amd64

What problem are you having? What should be the expected behaviour?

Amaru crashes when a rollback occurs

If possible, provide reproductions steps.

1. Run Amaru on preprod network
2. Wait for a rollback
3. Observe the node stalls and then crashes

If any, provide logs / error outputs from the system when the problem occurs.

{"timestamp":"2025-02-07T16:44:10.442129Z","level":"INFO","fields":{"message":"exit"},"target":"amaru_consensus::consensus","span":{"fork.length":2383,"peer":"preprod-node.world.dev.cardano.org:30000","rollback_point.slot":83199062,"name":"switch_to_fork"},"spans":[{"peer":"Peer { name: \"127.0.0.1:3001\" }","rollback":"(83263424, d1f71fcf3203a6addd9e87b
fdf65983f7e139fc63ac7972830409a68e73e9931)","name":"handle_roll_back"}]}                                                                                                          2025-02-07T16:44:10.442151Z  INFO switch_to_fork: amaru_consensus::consensus: exit peer="preprod-node.world.dev.cardano.org:30000" rollback_point.slot=83199062 fork.length=2383
{"timestamp":"2025-02-07T16:44:10.442162Z","level":"TRACE","fields":{"message":"exit"},"target":"amaru_consensus::consensus","span":{"peer":"Peer { name: \"127.0.0.1:3001\" }","r
ollback":"(83263424, d1f71fcf3203a6addd9e87bfdf65983f7e139fc63ac7972830409a68e73e9931)","name":"handle_roll_back"},"spans":[]}                                                    {"timestamp":"2025-02-07T16:44:10.442861Z","level":"TRACE","fields":{"message":"enter"},"target":"amaru::ledger","span":{"name":"forward"},"spans":[{"peer":"127.0.0.1:3001","name
":"pull.roll_back"},{"name":"forward"}]}
{"timestamp":"2025-02-07T16:44:10.442908Z","level":"TRACE","fields":{"message":"enter"},"target":"amaru_ledger","span":{"block.size":2471,"name":"parse_block"},"spans":[{"peer":"
127.0.0.1:3001","name":"pull.roll_back"},{"name":"forward"},{"block.size":2471,"name":"parse_block"}]}
{"timestamp":"2025-02-07T16:44:10.443060Z","level":"TRACE","fields":{"message":"exit"},"target":"amaru_ledger","span":{"block.size":2471,"name":"parse_block"},"spans":[{"peer":"1
27.0.0.1:3001","name":"pull.roll_back"},{"name":"forward"}]}
{"timestamp":"2025-02-07T16:44:10.443126Z","level":"TRACE","fields":{"message":"enter"},"target":"amaru::ledger::state","span":{"block.transactions.failed":0,"block.transactions.
success":1,"block.transactions.total":1,"name":"block.body.validate"},"spans":[{"peer":"127.0.0.1:3001","name":"pull.roll_back"},{"header.hash":"8e96380ec131c78540e05acc3d08707b9
24f07d1bafa22cb915da7b1887a0f4e","header.height":3164863,"header.slot":83199077,"name":"forward"},{"block.transactions.failed":0,"block.transactions.success":1,"block.transaction
s.total":1,"name":"block.body.validate"}]}

   amaru::fatal::error
   Whoops! The Amaru process panicked, rather than handling the error it encountered gracefully.

   This is almost certainly a bug, and we'd appreciate a report so we can improve Amaru.

   Please report this error at https://github.com/pragma-org/amaru/issues/new.

   In your bug report please provide the information below and if possible the code
   that produced it.

   Operating System: linux
   Architecture:     x86_64
   Version:          v0.1.0+dfe149a

   /home/curry/amaru/crates/ledger/src/state.rs:294:29

       unknown UTxO expected to be known: TransactionInput { transaction_id: Hash<32>("1022da6c73ef35976319889e64a707c3be6d8bff62a35f16c59d5a0fb7f29f35"), index: 2 }!

The salient entry is:

2025-02-07T16:44:10.442151Z  INFO switch_to_fork: amaru_consensus::consensus: exit peer="preprod-node.world.dev.cardano.org:30000" rollback_point.slot=83199062 fork.length=2383

which is a good hint about the origin of the bug. In the ChainSelector code we never reset the initial anchor which is used to compute where to rollback and then how many blocks to forward.

Please confirm that you've already searched issues, CHANGELOG and available documentation.

• I searched, but no luck.