lighthouse/book/src/merge-migration.md
Michael Sproul 1c9ec42dcb More merge doc updates (#3509)
## Proposed Changes

Address a few shortcomings of the book noticed by users:

- Remove description of redundant execution nodes
- Use an Infura eth1 node rather than an eth2 node in the merge migration example
- Add an example of the fee recipient address format (we support addresses without the 0x prefix, but 0x prefixed feels more canonical).
- Clarify that Windows support is no longer beta
- Add a link to the MSRV to the build-from-source instructions
2022-08-26 21:47:50 +00:00

11 KiB

Merge Migration

This document provides detail for users who want to run a merge-ready Lighthouse node.

The merge is occuring on mainnet in September. You must have a merge-ready setup by September 6 2022.

Necessary Configuration

There are two configuration changes required for a Lighthouse node to operate correctly throughout the merge:

  1. You must run your own execution engine such as Geth or Nethermind alongside Lighthouse. You must update your lighthouse bn configuration to connect to the execution engine using new flags which are documented on this page in the Connecting to an execution engine section.
  2. If your Lighthouse node has validators attached you must nominate an Ethereum address to receive transactions tips from blocks proposed by your validators. These changes should be made to your lighthouse vc configuration, and are covered on the Suggested fee recipient page.

Additionally, you must update Lighthouse to v3.0.0 (or later), and must update your execution engine to a merge-ready version.

When?

You must configure your node to be merge-ready before the Bellatrix fork occurs on the network on which your node is operating.

  • Mainnet: the Bellatrix fork is scheduled for epoch 144896, September 6 2022 11:34 UTC. You must ensure your node configuration is updated before then in order to continue following the chain. We recommend updating your configuration now.

  • Goerli (Prater), Ropsten, Sepolia, Kiln: the Bellatrix fork has already occurred. You must have a merge-ready configuration right now.

Connecting to an execution engine

The Lighthouse beacon node must connect to an execution engine in order to validate the transactions present in post-merge blocks. Two new flags are used to configure this connection:

  • --execution-endpoint <URL>: the URL of the execution engine API. Often this will be http://localhost:8551.
  • --execution-jwt <FILE>: the path to the file containing the JWT secret shared by Lighthouse and the execution engine.

If you set up an execution engine with --execution-endpoint then you must provide a JWT secret using --execution-jwt. This is a mandatory form of authentication that ensures that Lighthouse has authority to control the execution engine.

The execution engine connection must be exclusive, i.e. you must have one execution node per beacon node. The reason for this is that the beacon node controls the execution node. Please see the FAQ for further information about why many:1 and 1:many configurations are not supported.

Execution engine configuration

Each execution engine has its own flags for configuring the engine API and JWT. Please consult the relevant page for your execution engine for the required flags:

Once you have configured your execution engine to open up the engine API (usually on port 8551) you should add the URL to your lighthouse bn flags with --execution-endpoint <URL>, as well as the path to the JWT secret with --execution-jwt <FILE>.

There are merge-ready releases of all compatible execution engines available now.

Example

Let us look at an example of the command line arguments for a pre-merge production staking BN:

lighthouse \
    --network mainnet \
    beacon_node \
    --http \
    --eth1-endpoints http://localhost:8545,https://mainnet.infura.io/v3/TOKEN

Converting the above to a post-merge configuration would render:

lighthouse \
    --network mainnet \
    beacon_node \
    --http \
    --execution-endpoint http://localhost:8551
    --execution-jwt ~/.ethereum/geth/jwtsecret

The changes here are:

  1. Remove --eth1-endpoints
    • The endpoint at localhost can be retained, it is our local execution engine. Once it is upgraded to a merge-compatible release it will be used in the post-merge environment.
    • The infura.io endpoint will be abandoned, Infura and most other third-party node providers are not compatible with post-merge BNs.
  2. Add the --execution-endpoint flag.
    • We have reused the node at localhost, however we've switched to the authenticated engine API port 8551. All execution engines will have a specific port for this API, however it might not be 8551, see their documentation for details.
  3. Add the --execution-jwt flag.
    • This is the path to a file containing a 32-byte secret for authenticating the BN with the execution engine. In this example our execution engine is Geth, so we've chosen the default location for Geth. Your execution engine might have a different path. It is critical that both the BN and execution engine reference a file with the same value, otherwise they'll fail to communicate.

Note that the --network and --http flags haven't changed. The only changes required for the merge are ensuring that --execution-endpoint and --execution-jwt flags are provided! In fact, you can even leave the --eth1-endpoints flag there, it will be ignored. This is not recommended as a deprecation warning will be logged and Lighthouse may remove these flags in the future.

The relationship between --eth1-endpoints and --execution-endpoint

Pre-merge users will be familiar with the --eth1-endpoints flag. This provides a list of Ethereum "eth1" nodes (e.g., Geth, Nethermind, etc). Each beacon node (BN) can have multiple eth1 endpoints and each eth1 endpoint can have many BNs connection (many-to-many relationship). The eth1 node provides a source of truth for the deposit contract and beacon chain proposers include this information in beacon blocks in order to on-board new validators. BNs exclusively use the eth namespace on the eth1 JSON-RPC API to achieve this.

To progress through the Bellatrix upgrade nodes will need a new connection to an "eth1" node; --execution-endpoint. This connection has a few different properties. Firstly, the term "eth1 node" has been deprecated and replaced with "execution engine". Whilst "eth1 node" and "execution engine" still refer to the same projects (Geth, Nethermind, etc) the former refers to the pre-merge versions and the latter refers to post-merge versions. Secondly, there is a strict one-to-one relationship between Lighthouse and the execution engine; only one Lighthouse node can connect to one execution engine. Thirdly, it is impossible to fully verify the post-merge chain without an execution engine. It was possible to verify the pre-merge chain without an eth1 node, it was just impossible to reliably propose blocks without it.

Since an execution engine is a hard requirement in the post-merge chain and the execution engine contains the transaction history of the Ethereum chain, there is no longer a need for the --eth1-endpoints flag for information about the deposit contract. The --execution-endpoint can be used for all such queries. Therefore we can say that where --execution-endpoint is included --eth1-endpoints should be omitted.

FAQ

How do I know if my node is set up correctly?

Lighthouse will log a message indicating that it is ready for the merge:

INFO Ready for the merge, current_difficulty: 10789363, terminal_total_difficulty: 10790000

Once the merge has occurred you should see that Lighthouse remains in sync and marks blocks as verified indicating that they have been processed successfully by the execution engine:

INFO Synced, slot: 3690668, block: 0x1244…cb92, epoch: 115333, finalized_epoch: 115331, finalized_root: 0x0764…2a3d, exec_hash: 0x929c…1ff6 (verified), peers: 78

Can I still use the --staking flag?

Yes. The --staking flag is just an alias for --http --eth1. The --eth1 flag is now superfluous so --staking is equivalent to --http. You need either --staking or --http for the validator client to be able to connect to the beacon node.

Can I use http://localhost:8545 for the execution endpoint?

Most execution nodes use port 8545 for the Ethereum JSON-RPC API. Unless custom configuration is used, an execution node will not provide the necessary engine API on port 8545. You should not attempt to use http://localhost:8545 as your engine URL and should instead use http://localhost:8551.

Can I share an execution node between multiple beacon nodes (many:1)?

It is not possible to connect more than one beacon node to the same execution engine. There must be a 1:1 relationship between beacon nodes and execution nodes.

The beacon node controls the execution node via the engine API, telling it which block is the current head of the chain. If multiple beacon nodes were to connect to a single execution node they could set conflicting head blocks, leading to frequent re-orgs on the execution node.

We imagine that in future there will be HTTP proxies available which allow users to nominate a single controlling beacon node, while allowing consistent updates from other beacon nodes.

What about multiple execution endpoints (1:many)?

It is not possible to connect one beacon node to more than one execution engine. There must be a 1:1 relationship between beacon nodes and execution nodes.

Since an execution engine can only have one controlling BN, the value of having multiple execution engines connected to the same BN is very low. An execution engine cannot be shared between BNs to reduce costs.

Whilst having multiple execution engines connected to a single BN might be useful for advanced testing scenarios, Lighthouse (and other consensus clients) have decided to support only one execution endpoint. Such scenarios could be resolved with a custom-made HTTP proxy.

Additional Resources

There are several community-maintained guides which provide more background information, as well as guidance for specific setups.