Run a cluster with others

The following instructions aim to assist a group of users coordinating together to create a distributed validator cluster between them. Only one person needs to do step 2 and step 5 in the quickstart process.

Pre-requisites

Ensure you have docker and git installed. Also, make sure docker is running before executing the commands below.

Step 1. Creating and backing up a private key for charon

The first step of running a cluster is preparing for a distributed key generation ceremony. To do this everyone must create an ENR for their charon client. This ENR is a public/private key pair, and allows the other charon clients in the DKG to identify and connect to your node.

# Clone this repo
git clone https://github.com/ObolNetwork/charon-distributed-validator-node.git

# Change directory
cd charon-distributed-validator-node

# Create your charon ENR private key, this will create a charon-enr-private-key file in the .charon directory
docker run --rm -v "$(pwd):/opt/charon" obolnetwork/charon:v0.10.0 create enr

You should expect to see a console output like

Created ENR private key: .charon/charon-enr-private-key enr:-JG4QGQpV4qYe32QFUAbY1UyGNtNcrVMip83cvJRhw1brMslPeyELIz3q6dsZ7GblVaCjL_8FKQhF6Syg-O_kIWztimGAYHY5EvPgmlkgnY0gmlwhH8AAAGJc2VjcDI1NmsxoQKzMe_GFPpSqtnYl-mJr8uZAUtmkqccsAx7ojGmFy-FY4N0Y3CCDhqDdWRwgg4u

⚠️ Attention

Please make sure to create a backup of the private key at .charon/charon-enr-private-key. Be careful not to commit it to git! If you lose this file you won't be able to take part in the DKG ceremony.

If you are taking part in an organised Obol testnet, submit the created ENR public address (the console output starting with enr:-... not the contents of the private key file) to the appropriate typeform.

Step 2. Leader creates the DKG configuration file and distributes it to everyone else

One person, in the cluster or otherwise, will prepare the configuration file for the distributed key generation ceremony using the charon create dkg command. For the official Obol testnets, this step will be completed by an Obol core team member or the cluster captain and the definition file will be distributed to the cluster members for DKG completion.

In future, step 1 and step 2 of this guide will use the Obol Distributed Validator Launchpad to facilitate and verify these files are created in an authenticated manner.

# Prepare an environment variable file
cp .env.sample .env

# Set the ENRs of all the operators participating in the DKG ceremony in the .env file variable CHARON_OPERATOR_ENRS

# Set FEE_RECIPIENT_ADDRESS and WITHDRAWAL_ADDRESS to ETH1 addresses of your choice.
# NAME can be any random string like "Obol Team"
docker run --rm -v "$(pwd):/opt/charon" --env-file .env obolnetwork/charon:v0.10.0 create dkg --name=$NAME --fee-recipient-address=$FEE_RECIPIENT_ADDRESS --withdrawal-address=$WITHDRAWAL_ADDRESS

# The above command prepares a DKG configuration file.

This command should output a file at .charon/cluster-definition.json. This file needs to be shared with the other operators in a cluster.

Step 3. Run the DKG

After receiving the cluster-definition.json file created by the leader, cluster members should ideally save it in the .charon/ folder that was created during step 1, alternatively the --definition-file flag can override the default expected location for this file.

Every cluster member then participates in the DKG ceremony. For Charon v1, this needs to happen synchronously between participants at an agreed time.

# Participate in DKG ceremony, this will create .charon/cluster-lock.json, .charon/deposit-data.json and .charon/validator_keys
docker run --rm -v "$(pwd):/opt/charon" obolnetwork/charon:v0.10.0 dkg --p2p-bootnode-relay

Assuming the DKG is successful, a number of artefacts will be created in the .charon folder. These include:

• A deposit-data.json file. This contains the information needed to activate the validator on the Ethereum network.
• A cluster-lock.json file. This contains the information needed by charon to operate the distributed validator cluster with its peers.
• A validator_keys/ folder. This folder contains the private key shares and passwords for the created distributed validators.

At this point you should make a backup of the .charon/validator_keys folder as replacing lost private keys is not straightforward at this point in charon's development. The cluster-lock and deposit-data files are identical for each operator and can be copied if lost.

If taking part in the official Athena testnet, one cluster member will have to submit the cluster-lock and deposit-data files to the Obol Team, setting the stage for activation.

Step 4. Start the Distributed Validator Cluster

With the DKG ceremony over, the last phase before activation is to prepare your node for validating over the long term. This repo is configured to sync an execution layer client (geth) and a consensus layer client (lighthouse).

Before completing these instructions, you should assign a static local IP address to your device (extending the DHCP reservation indefinitely or removing the device from the DCHP pool entirely if you prefer), and port forward the TCP protocol on the public port :3610 on your router to your device's local IP address on the same port. This step is different for every person's home internet, and can be complicated by the presence of dynamic public IP addresses. We are currently working on making this as easy as possible, but for the time being, a distributed validator cluster isn't going to work very resiliently if all charon nodes cannot talk directly to one another and instead need to have an intermediary node forwarding traffic to them.

Caution: If you manually update docker-compose to mount lighthouse from your locally synced ~/.lighthouse, the whole chain database may get deleted. It'd be best not to manually update as lighthouse checkpoint-syncs so the syncing doesn't take much time.

NOTE: If you have a geth node already synced, you can simply copy over the directory. For ex: cp -r ~/.ethereum/goerli data/geth. This makes everything faster since you start from a synced geth node.

# Delete lighthouse data if it exists
rm -r ./data/lighthouse

# Spin up a Distributed Validator Node with a Validator Client
docker-compose up

# Open Grafana dashboard
open http://localhost:3000/d/singlenode/

You should use the grafana dashboard to infer whether your cluster is healthy. In particular you should check:

• That your charon client can connect to the configured beacon client.
• That your charon client can connect to all peers

You might notice that there are logs indicating that a validator cannot be found and that APIs are returning 404. This is to be expected at this point, as the validator public keys listed in the lock file have not been deposited and acknowledged on the consensus layer yet (usually ~16 hours after the deposit is made).

To turn off your node after checking the health of the cluster you can run:

# Shut down the currently running distributed validator node
docker-compose down

Step 5. Activate the deposit data

If you and your team have gotten to this phase of the quickstart, and you have successfully created a distributed validator together, and you have connected all of your charon clients together such that the monitoring indicates that they are all healthy and ready to operate, one person may process to activate this deposit data with the existing staking launchpad.

This process can take a minimum of 16 hours, with the maximum time to activation being dictated by the length of the activation queue, which can be weeks. You can leave your distributed validator cluster offline until closer to the activation period if you would prefer. You can also use this time to improve and harden your monitoring and alerting for the cluster.

If you have gotten this far through the process, and whether you succeed or fail at running the distributed validator successfully on the testnet, we would like to hear your feedback on the process and where you encountered difficulties. Please open issues in either this repo if the problem is deployment related, or the charon repo if the issue is directly related to the client.

Other Actions

The above steps should get you running a distributed validator cluster. The following are some extra steps you may want to take either to help Obol with their testing program, or to improve the resilience and performance of your distributed validator cluster.

Leader Adds Central Monitoring Token

The cluster leader will be provided with a Central Monitoring Token used to push distributed validator metrics to our central prometheus service to monitor, analyze and improve your cluster's performance. The token needs to be added in prometheus/prometheus.yml replacing $PROM_REMOTE_WRITE_TOKEN. The token will look like: eyJtZXNzYWdlIjoiSldUIFJ1bGVzISIsImlhdCI6MTQ1OTQ0ODExOSwiZXhwIjoxNDU5NDU0NTE5fQ. Final prometheus/prometheus.yml would look something like:

global:
  scrape_interval:     12s # Set the scrape interval to every 12 seconds. Default is every 1 minute.
  evaluation_interval: 12s # Evaluate rules every 12 seconds. The default is every 1 minute.

remote_write:
  - url: https://prometheus-prod-10-prod-us-central-0.grafana.net/api/prom/push
    authorization:
      credentials: 436764:eyJtZXNzYWdlIjoiSldUIFJ1bGVzISIsImlhdCI6MTQ1OTQ0ODExOSwiZXhwIjoxNDU5NDU0NTE5fQ
    name: obol-prom

scrape_configs:
  - job_name: 'charon'
    static_configs:
      - targets: ['charon:3620']
  - job_name: 'teku'
    static_configs:
      - targets: ['teku:8008']
  - job_name: 'node-exporter'
    static_configs:
      - targets: ['node-exporter:9100']

Validator Voluntary Exit

A voluntary exit is when a validator chooses to stop performing its duties, and exits the beacon chain permanently. To voluntarily exit, the validator must continue performing its validator duties until successfully exited to avoid penalties.

To trigger a voluntary exit, a sidecar docker-compose command is executed that signs and submits the voluntary exit to the active running charon node that shares it with other nodes in the cluster. The commands below should be executed on the same machine and same folder as the active running charon-distribute-validator-node docker compose.

info

A threshold of peers in the cluster need to perform this task to exit a validator.

• Create a new exit_keys folder next to .charon/validator_keys: mkdir .charon/exit_keys
• Copy the validator keys and passwords that you want to exit from the validator_keys folder to the exit_keys folder.
  • E.g. to exit validator #4: cp .charon/validator_keys/keystore/keystore-4* .charon/exit_keys/
  • Warning: all keys copied to the exit_keys folder will be exited, so be careful!
• Ensure the external network in compose-voluntary-exit.yml is correct.
  • Confirm the name of the exiting charon-distributed-validator-node docker network: docker network ls.
  • If it isn't charon-distributed-validator-node-dvnode, then update compose-voluntary-exit.yml accordingly.
• Ensure the latest fork version epoch is used:
  • Voluntary exists require an epoch after which they take effect.
  • All VCs need to sign and submit the exact same messages (epoch) in DVT.
  • --epoch=1 would be ideal, since all chains have that epoch in the past, so the validator should exit immediately.
  • There is however a bug in lighthouse requiring an epoch that maps to the latest fork version to be used.
  • compose-voluntary-exit.yml is configured with --epoch=112260 which is the latest Bellatrix fork on Prater.
  • If the Charon cluster is running on a different chain, ALL operators must update --epoch to the same latest fork version returned by curl $BEACON_NODE/eth/v1/config/fork_schedule.
• Run the command to submit this node's partially signed voluntary exit:
  • docker-compose -f compose-voluntary-exit.yml up
  • Confirm the logs: Exit for validator XXXXX submitted
  • Exit the container: Ctrl-C
• The charon metric core_parsigdb_exit_total will be incremented each time a voluntary exit partial signature is received, either from this node or from peers.

Self-Host a Bootnode

If you are experiencing connectivity issues with the Obol hosted bootnode, or you want to improve your clusters latency and decentralisation, you can opt to host your own bootnode on a separate open and static internet port.

# Figure out your public IP
curl v4.ident.me

# Clone the repo and cd into it.
git clone https://github.com/ObolNetwork/charon-distributed-validator-node.git

cd charon-distributed-validator-node

# Replace 'replace.with.public.ip.or.hostname' in bootnode/docker-compose.yml with your public IPv4 or DNS hostname # Replace 'replace.with.public.ip.or.hostname' in bootnode/docker-compose.yml with your public IPv4 or DNS hostname

nano bootnode/docker-compose.yml

docker-compose -f bootnode/docker-compose.yml up

Test whether the bootnode is publicly accessible. This should return an ENR: curl http://replace.with.public.ip.or.hostname:3640/enr

Ensure the ENR returned by the bootnode contains the correct public IP and port by decoding it with https://enr-viewer.com/.

Configure ALL charon nodes in your cluster to use this bootnode:

• Either by adding a flag: --p2p-bootnodes=http://replace.with.public.ip.or.hostname:3640/enr
• Or by setting the environment variable: CHARON_P2P_BOOTNODES=http://replace.with.public.ip.or.hostname:3640/enr

Note that a local boonode/.charon/charon-enr-private-key file will be created next to bootnode/docker-compose.yml to ensure a persisted bootnode ENR across restarts.