Experimental --Xchain-pruning-enabled=true doesn't work with documented QBFT defaults

[matthew1001]

When a Bonsai QBFT node with --Xchain-pruning-enabled=true is stopped and restarted, it errors with:

2024-11-29 15:26:30.120+00:00 | main | ERROR | Runner | unable to start main loop
com.google.common.util.concurrent.UncheckedExecutionException: java.util.NoSuchElementException: Supplied block was on a orphaned chain, unable to generate VoteTally.
        at com.google.common.cache.LocalCache$Segment.get(LocalCache.java:2085)
        at com.google.common.cache.LocalCache.get(LocalCache.java:4017)
        at com.google.common.cache.LocalCache$LocalManualCache.get(LocalCache.java:4898)
        at org.hyperledger.besu.consensus.common.validator.blockbased.VoteTallyCache.getVoteTallyAfterBlock(VoteTallyCache.java:76)
        at org.hyperledger.besu.consensus.common.validator.blockbased.BlockValidatorProvider.getValidatorsAfterBlock(BlockValidatorProvider.java:97)
        at org.hyperledger.besu.consensus.qbft.validator.ForkingValidatorProvider.getValidatorsAfterBlock(ForkingValidatorProvider.java:66)
        at org.hyperledger.besu.consensus.qbft.validator.ForkingValidatorProvider.getValidatorsAtHead(ForkingValidatorProvider.java:60)
        at org.hyperledger.besu.consensus.common.bft.statemachine.BftFinalState.getValidators(BftFinalState.java:93)
        at org.hyperledger.besu.consensus.common.bft.statemachine.BftFinalState.isLocalNodeValidator(BftFinalState.java:130)
        at org.hyperledger.besu.consensus.qbft.statemachine.QbftBlockHeightManagerFactory.create(QbftBlockHeightManagerFactory.java:69)
        at org.hyperledger.besu.consensus.qbft.statemachine.QbftController.createNewHeightManager(QbftController.java:118)
        at org.hyperledger.besu.consensus.common.bft.statemachine.BaseBftController.startNewHeightManager(BaseBftController.java:210)
        at org.hyperledger.besu.consensus.common.bft.statemachine.BaseBftController.start(BaseBftController.java:79)
        at org.hyperledger.besu.consensus.common.bft.blockcreation.BftMiningCoordinator.start(BftMiningCoordinator.java:101)
        at org.hyperledger.besu.controller.QbftBesuControllerBuilder$1.onInitialSyncCompleted(QbftBesuControllerBuilder.java:323)

Steps to Reproduce

1. Create 4-node QBFT chain
   • I used a mix of FOREST and BONSAI nodes, all with FULL sync mode
2. Configure one of the BONSAI nodes to have --Xchain-pruning-enabled
3. Start the chain (2s block period) and leave it for long enough to produce 7200 blocks (the min default for pruning)
4. At e.g. 10k blocks, stop all the nodes
5. Restart all the nodes
6. Nodes 1-3 start producing blocks again. Node 4 errors with above stack trace.

Version

24.11-develop-5596cbd

[matthew1001]

It looks like this is due to the default epochlength in the docs being set to 30000 blocks, where pruning only retains the previous 7200 blocks. So changing config does fix the issue, but I think the flag should check if epochlength > --Xchain-pruning-blocks-retained and fail validation if it does. I'll leave this issue open and will maybe find time to fix it shortly.