Live tracing experimental MVP support for arbitrum nitro #2934
Conversation
|
|
| } | ||
|
|
||
| defer func() { | ||
| getVMConfig().Tracer.OnBlockEnd(outError) |
There was a problem hiding this comment.
This hook was missing a call, so I've added it manually here
| @@ -192,6 +218,16 @@ func ProduceBlockAdvanced( | |||
| } | |||
|
|
|||
| header := createNewHeader(lastBlockHeader, l1Info, arbState, chainConfig) | |||
|
|
|||
| getVMConfig().Tracer.OnBlockStart( | |||
There was a problem hiding this comment.
This hook was missing a call, so I've added it manually here
| [submodule "go-ethereum"] | ||
| path = go-ethereum | ||
| url = https://github.com/Tenderly/arb-go-ethereum.git | ||
| branch = live-tracing |
There was a problem hiding this comment.
remove this change.
You should have a separate PR to our go-ethereum fork, point to it from your comment ("this pulls in + link"), and add a commit which sets go-ethereum to the specific commit of your geth-change.
We have merge checks that will not allow nitro commit to be merged before the geth PR is also merged.
| cacheConfig *core.CacheConfig, | ||
| chainConfig *params.ChainConfig, | ||
| txLookupLimit uint64, | ||
| tracingConfig json.RawMessage, |
There was a problem hiding this comment.
I think you should not need to chainge blockchain at all. However, if you do - seems better to get here *tracing.Hooks that a json.RawMessage.
| } | ||
|
|
||
| // TODO: IMPORTANT !!! | ||
| // This will cause a tracer to be called multiple times for the same block concurrently |
There was a problem hiding this comment.
This is probably not the only thing that'll cause tracer to be called multiple times.
In normal case we have no reorgs, but a node might reexecute the same block multiple times for ultiple reasons.
I didn't view geth side yet, but I think a better solution will be to:
- remove all changes to blockchain
- Pass a vm.Config to ProduceBlock which could enable tracing
- have a config that controls tracing and feed it to ExecutionEngine (it currently does not have a config.. have a look how we use configs for e.g. SyncMonitorConfig)
- Inside func (s *ExecutionEngine) createBlockFromNextMessage - pass to ProduceBlock a vm.Config that enables tracing if it was enabled in config and only if isMsgForPrefetch == false
| @@ -172,7 +184,21 @@ func ProduceBlockAdvanced( | |||
| sequencingHooks *SequencingHooks, | |||
| isMsgForPrefetch bool, | |||
| runMode core.MessageRunMode, | |||
| ) (*types.Block, types.Receipts, error) { | |||
| blockChain *core.BlockChain, | |||
There was a problem hiding this comment.
seems better to get a VMConfig than a BlockChain.
Integrating Tenderly Live Tracer into Arbitrum Nitro
Description:
This PR demonstrates our attempt to integrate live tracing into Arbitrum Nitro.
While we have a more complex system in place with the actual live tracer and a full build system around it, this PR highlights the core changes necessary to enable live tracing for Arb Nitro.
It serves as a foundation for understanding the key modifications required for this integration to happen.