Validator Tasks

This guide focuses on running the Sui node software as a validator.

Requirements

To run a Sui Validator a machine with the following is required:

• CPU: 24 physical cores (or 48 virtual cores)
• Memory: 128 GB
• Storage: 4 TB NVME
• Network: 1 Gbps

Deployment

You can deploy Sui node in a number of ways.

There are pre-built container images available in Docker Hub.

And pre built linux/amd64 binaries available in S3 that you can can fetch using one of the following methods:

wget https://releases.sui.io/$SUI_SHA/sui-node

curl https://releases.sui.io/$SUI_SHA/sui-node -o sui-node

To build directly from source:

git clone https://github.com/MystenLabs/sui.git && cd sui
git checkout [SHA|BRANCH|TAG]
cargo build --release --bin sui-node

Configuration and guides are available for the following deployment options:

• Systemd
• Ansible
• Docker Compose

Configuration

Sui node runs with a single configuration file provided as an argument, example:

./sui-node --config-path /opt/sui/config/validator.yaml.

See Validator for configuration templates.

Connectivity

Sui node uses the following ports by default:

protocol/port	reachability	purpose
TCP/8080	inbound	protocol/transaction interface
UDP/8081	inbound/outbound	narwhal primary interface
UDP/8082	inbound/outbound	narwhal worker interface
TCP/8083	localhost	sui -> narwhal interface
UDP/8084	inbound/outbound	peer to peer state sync interface
TCP/8443	outbound	metrics pushing
TCP/9184	localhost	metrics scraping

To run a validator successfully, it is critical that ports 8080-8084 are open as outlined, including the specific protocol (TCP/UDP).

Network Buffer

From load testing Sui validator networks, it has been determined that the default Linux network buffer sizes are too small. We recommend increasing them using one of the following two methods:

Option 1: With /etc/sysctl.d/

These settings can be added to a new sysctl file specifically for the sui-node, or appended to an existing file. Modifications made in this way will persist across system restarts.

# Create a new sysctl file for the sui-node
sudo nano /etc/sysctl.d/100-sui-node.conf

# Add these lines to the file, overwriting existing settings if necessary.
net.core.rmem_max = 104857600
net.core.wmem_max = 104857600
net.ipv4.tcp_rmem = 8192 262144 104857600
net.ipv4.tcp_wmem = 8192 262144 104857600

# Apply the settings immediately, before the next restart
sudo sysctl --system

Option 2: With sysctl command

These modifications do not persist across system restarts. Therefore, the commands should be run each time the host restarts.

sudo sysctl -w net.core.wmem_max=104857600
sudo sysctl -w net.core.rmem_max=104857600
sudo sysctl -w net.ipv4.tcp_rmem="8192 262144 104857600"
sudo sysctl -w net.ipv4.tcp_wmem="8192 262144 104857600"

Verification

To verify that the system settings have indeed been updated, check the output of the following command:

sudo sysctl -a | egrep [rw]mem

Storage

All Sui node related data is stored by default under /opt/sui/db/. This is controlled in the Sui node configuration file.

$ cat /opt/sui/config/validator.yaml | grep db-path
  db-path: /opt/sui/db/authorities_db
  db-path: /opt/sui/db/consensus_db

Ensure that you have an appropriately sized disk mounted for the database to write to.

• To check the size of the local Sui node databases:

du -sh /opt/sui/db/
du -sh /opt/sui/db/authorities_db
du -sh /opt/sui/db/consensus_db

• To delete the local Sui node databases:

sudo systemctl stop sui-node
sudo rm -rf /opt/sui/db/authorities_db /opt/sui/db/consensus_db

Key management

The following keys are used by Sui node:

key	scheme	purpose
protocol.key	bls12381	transactions, narwhal consensus
account.key	ed25519	controls assets for staking
network.key	ed25519	narwhal primary, sui state sync
worker.key	ed25519	validate narwhal workers

These are configured in the Sui node configuration file.

Monitoring

Metrics

Sui node exposes metrics via a local HTTP interface. These can be scraped for use in a central monitoring system as well as viewed directly from the node.

• View all metrics:

curl -s http://localhost:9184/metrics

• Search for a particular metric:

curl http://localhost:9184/metrics | grep <METRIC>

Sui node also pushes metrics to a central Sui metrics proxy.

Logs

Logs are controlled using the RUST_LOG environment variable.

The RUST_LOG_JSON=1 environment variable can optionally be set to enable logging in JSON structured format.

Depending on your deployment method, these are configured in the following places:

• Ansible
• Native systemd
• Docker Compose

To view and follow the Sui node logs:

journalctl -u sui-node -f

To search for a particular match

journalctl -u sui-node -g <SEARCH_TERM>

• If using Docker Compose, look at the examples in the README.

It is possible to change the logging configuration while a node is running using the admin interface.

To view the currently configured logging values:

curl localhost:1337/logging

To change the currently configured logging values:

curl localhost:1337/logging -d "info"

Dashboards

Public dashboard for network wide visibility:

• Sui Testnet Validators

Software updates

When an update is required to the Sui node software the following process can be used. Follow the relevant Systemd or Docker Compose runbook depending on your deployment type. It is highly unlikely that you will want to restart with a clean database.

• Systemd
• Docker Compose

State sync

Checkpoints in Sui contain the permanent history of the network. They are comparable to blocks in other blockchains with one big difference being that they are lagging instead of leading. All transactions are final and executed prior to being included in a checkpoint.

These checkpoints are synchronized between validators and fullnodes via a dedicated peer to peer state sync interface.

Inter-validator state sync is always permitted however there are controls available to limit what fullnodes are allowed to sync from a specific validator.

The default and recommended max-concurrent-connections: 0 configuration does not affect inter-validator state sync, but will restrict all fullnodes from syncing. The Sui node configuration can be modified to allow a known fullnode to sync from a validator:

p2p-config:
  anemo-config:
    max-concurrent-connections: 0
  seed-peers:
    - address: <multiaddr>  # The p2p address of the fullnode
      peer-id: <peer-id>    # hex encoded network public key of the node
    - address: ...          # another permitted peer
      peer-id: ...

Chain operations

The following chain operations are executed using the sui CLI. This binary is built and provided as a release similar to sui-node, examples:

wget https://releases.sui.io/$SUI_SHA/sui
chmod +x sui

curl https://releases.sui.io/$SUI_SHA/sui -o sui
chmod +x sui

It is recommended and often required that the sui binary release/version matches that of the deployed network.

Updating on-chain metadata

You can leverage Validator Tool to perform majority of the following tasks.

An active/pending validator can update its on-chain metadata by submitting a transaction. Some metadata changes take effect immediately, including:

• name
• description
• image url
• project url

Other metadata (keys, addresses, and so on) only come into effect at the next epoch.

To update metadata, a validator makes a MoveCall transaction that interacts with the System Object. For example:

1. to update name to new_validator_name, use the Sui Client CLI to call sui_system::update_validator_name:

tip

Beginning with the Sui v1.24.1 release, the --gas-budget flag is no longer required for CLI commands.

sui client call --package 0x3 --module sui_system --function update_validator_name --args 0x5 \"new_validator_name\" --gas-budget 10000

2. to update p2p address starting from next epoch to /ip4/192.168.1.1, use the Sui Client CLI to call sui_system::update_validator_next_epoch_p2p_address:

sui client call --package 0x3 --module sui_system --function update_validator_next_epoch_p2p_address --args 0x5 "[4, 192, 168, 1, 1]" --gas-budget 10000

See the full list of metadata update functions here.

Operation cap

To avoid touching account keys too often and allowing them to be stored off-line, validators can delegate the operation ability to another address. This address can then update the reference gas price and tallying rule on behalf of the validator.

Upon creating a Validator, an UnverifiedValidatorOperationCap is created as well and transferred to the validator address. The holder of this Cap object (short for "Capability") therefore could perform operational actions for this validator. To authorize another address to conduct these operations, a validator transfers the object to another address that they control. The transfer can be done by using Sui Client CLI: sui client transfer.

To rotate the delegatee address or revoke the authorization, the current holder of Cap transfers it to another address. In the event of compromised or lost keys, the validator could create a new Cap object to invalidate the incumbent one. This is done by calling sui_system::rotate_operation_cap:

sui client call --package 0x3 --module sui_system --function rotate_operation_cap --args 0x5 --gas-budget 10000

By default the new Cap object is transferred to the validator address, which then could be transferred to the new delegatee address. At this point, the old Cap becomes invalidated and no longer represents eligibility.

To get the current valid Cap object's ID of a validator, use the Sui Client CLI sui client objects command after setting the holder as the active address.

Updating the gas price survey quote

To update the gas price survey quote of a validator, which is used to calculate the reference gas price at the end of the epoch, the sender needs to hold a valid UnverifiedValidatorOperationCap. The sender could be the validator itself, or a trusted delegatee. To do so, call sui_system::request_set_gas_price:

sui client call --package 0x3 --module sui_system --function request_set_gas_price --args 0x5 {cap_object_id} {new_gas_price} --gas-budget 10000

Reporting/un-reporting validators

To report a validator or undo an existing reporting, the sender needs to hold a valid UnverifiedValidatorOperationCap. The sender could be the validator itself, or a trusted delegatee. To do so, call sui_system::report_validator/undo_report_validator:

sui client call --package 0x3 --module sui_system --function report_validator/undo_report_validator --args 0x5 {cap_object_id} {reportee_address} --gas-budget 10000

After a validator is reported by 2f + 1 other validators by voting power, their staking rewards will be slashed.

Joining the validator set

In order for a Sui address to join the validator set, they need to first sign up as a validator candidate by calling sui_system::request_add_validator_candidate with their metadata and initial configs:

sui client call --package 0x3 --module sui_system --function request_add_validator_candidate --args 0x5 {protocol_pubkey_bytes} {network_pubkey_bytes} {worker_pubkey_bytes} {proof_of_possession} {name} {description} {image_url} {project_url} {net_address} {p2p_address} {primary_address} {worker_address} {gas_price} {commission_rate} --gas-budget 10000

After an address becomes a validator candidate, any address (including the candidate address itself) can start staking with the candidate's staking pool. Refer to our dedicated staking FAQ on how staking works. Once a candidate's staking pool has accumulated at least sui_system::MIN_VALIDATOR_JOINING_STAKE amount of stake, the candidate can call sui_system::request_add_validator to officially add themselves to next epoch's active validator set:

sui client call --package 0x3 --module sui_system --function request_add_validator --args 0x5 --gas-budget 10000000

Leaving the validator set

To leave the validator set starting next epoch, the sender needs to be an active validator in the current epoch and should call sui_system::request_remove_validator:

sui client call --package 0x3 --module sui_system --function request_remove_validator --args 0x5 --gas-budget 10000

After the validator is removed at the next epoch change, the staking pool will become inactive and stakes can only be withdrawn from an inactive pool.

Private security fixes

There might be instances where urgent security fixes need to be rolled out before publicly announcing it's presence (issues affecting liveliness, invariants such as SUI supply, governance, and so on). To not be actively exploited, Mysten Labs will release signed security binaries incorporating such fixes with a delay in publishing the source code until a large percentage of our validators have patched the vulnerability.

This release process is different and we expect us to announce the directory for such binaries out of band. Our public key to verify these binaries would be stored here

There is also a script available that downloads all the necessary signed binaries and docker artifacts incorporating the security fixes.

Usage ./download_private.sh <directory-name>

You can also download and verify specific binaries that may not be included by the above script using the download_and_verify_private_binary.sh script.

Usage: ./download_and_verify_private_binary.sh <directory-name> <binary-name>