Skip to main content

​
Overview

This tutorial teaches you how to rewind a single op-geth instance to a previous chain head. This can be helpful to fix a divergent node, recreate archival state, or the safedb from an earlier chain state.

​
Rewind op-geth

Follow these steps to rewind your op-geth node.
1

Shut down the CL

Whatever CL clients are driving the op-geth instance need to be stopped to prevent them from interfering with the rewind later. This would typically be a locally running op-node or kona-node application, a Docker container or a pod in a Kubernetes cluster.
Make sure you don’t have something like ArgoCD syncing the app in question which would just scale the pod back up again.
2

Determine rewind target

If you already know the block number/hash you want to rewind to, you can proceed to the next step.If you only know the timestamp you want to rewind to, you can find the corresponding block with
cast find-block 1741890000 -r <ETH_RPC_URL>
Where <ETH_RPC_URL> points to your op-geth node’s RPC endpoint, e.g. http://localhost:8545 if it was a local instance with the default port and 1741890000 is an example unix timestamp. You can use e.g. https://www.epochconverter.com/ to convert from human times to unix timestamps.
3

Rewind op-geth

You need the JWT secret and open & admin port (default 8545 & 8551). Then you can use the op-wheel command from the monorepo to issue the rewind.
# save JWT secret to env var
export JWT="YourJWTSecret123"
# assuming default ports 8545 and 8551 and local endpoint
go run ./op-wheel/cmd engine rewind \
--engine.open http://localhost:8545 \
--engine http://localhost:8551 \
--engine.jwt-secret $JWT \
--log.level DEBUG \
--set-head --to 8460000
This would rewind the op-geth node to the block with number 8460000. You should have determined the right block number in step 2.
4

Start the CL

You can now start your CL client again. It will sync from the rewound block head.