Sample AWS Blockchain Node Runner app for Solana Nodes

Contributed by
@frbrkoala

Solana nodes on AWS can be deployed in 3 different configurations: Consensus, base RPC and extended RPC with secondary indexes. In addition, you can choose to deploy those configurations as a single node or a highly available (HA) nodes setup. Learn more about configurations on Solana on AWS documentation page and below are the details on single node and HA deployment setups.

Overview of Deployment Architectures for Single and HA setups

Single node setup

Single Node Deployment

A Solana node deployed in the Default VPC continuously synchronizes with the rest of nodes on Solana Clusters through Internet Gateway.
The Solana node is used by dApps or development tools internally from within the Default VPC. JSON RPC API is not exposed to the Internet directly to protect nodes from unauthorized access.
The Solana node uses all required secrets locally, but stores a copy in AWS Secrets Manager as secure backup.
The Solana node sends various monitoring metrics for both EC2 and Solana nodes to Amazon CloudWatch.

HA setup

Highly Available Nodes Deployment

A set of Base or Extended RPC Solana nodes are deployed within the Auto Scaling Group in the Default VPC continuously synchronizes with the rest of nodes on Solana Clusters through Internet Gateway. Note that HA setup is not suitable for Consensus nodes.
The Solana nodes are accessed by dApps or development tools internally through Application Load Balancer. JSON RPC API is not exposed to the Internet to protect nodes from unauthorized access. dApps need to handle user authentication and API protection, like in this example for dApps on AWS.
The Solana nodes use all required secrets locally, but store a copy in AWS Secrets Manager as secure backup.
The Solana nodes send various monitoring metrics for both EC2 and Solana nodes to Amazon CloudWatch.

Additional materials

Managing Secrets

During the startup, if a node can't find the necessary identity file on the attached Root EBS volume, it generates a new one and stores it in AWS Secrets Manager. For a single-node deployment, the ARN of a secret can be provided within the .env configuration file with configuration and the node will pick it up.

Base RPC and Extended RPC nodes use only 1 secret:

Solana Node Identity Secret: The identity key pair for a Solana node.

Consensus node uses up to 3 more identity secrets:

Vote Account Secret: The Validator Identity's key pair.
Authorized Withdrawer Account Secret: The Authorized Withdrawer key pair.
Registration Transaction Funding Account Secret: An account that has sufficient SOL to pay for on-chain validator creation transaction. If not present, the node provisioning script assumes the on-chain validator creation transaction was issued elsewhere and will skip it.

Well-Architected Checklist

This is the Well-Architected checklist for Solana nodes implementation of the AWS Blockchain Node Runner app. This checklist takes into account questions from the AWS Well-Architected Framework which are relevant to this workload. Please feel free to add more checks from the framework if required for your workload.

Pillar	Control	Question/Check	Remarks
Security	Network protection	Are there unnecessary open ports in security groups?	Please note that ports 8801 to 8812 (TCP/UDP) for Solana are open to public to support P2P protocols. We have to rely on the protection mechanisms built into the Solana validators software to protect those ports.
		Traffic inspection	Traffic protection is not used in the solution. AWS Web Applications Firewall (WAF) could be implemented for traffic inspection. Additional charges will apply.
	Compute protection	Reduce attack surface	This solution uses Ubuntu 20.04 AMI. You may choose to run hardening scripts on it.
		Enable people to perform actions at a distance	This solution uses AWS Systems Manager for terminal session, not ssh ports.
	Data protection at rest	Use encrypted Amazon Elastic Block Store (Amazon EBS) volumes	This solution uses encrypted Amazon EBS volumes.
		Use encrypted Amazon Simple Storage Service (Amazon S3) buckets	This solution uses Amazon S3 managed keys (SSE-S3) encryption.
	Data protection in transit	Use TLS	The AWS Application Load balancer currently uses HTTP listener. Create HTTPS listener with self signed certificate if TLS is desired.
	Authorization and access control	Use instance profile with Amazon Elastic Compute Cloud (Amazon EC2) instances	This solution uses AWS Identity and Access Management (AWS IAM) role instead of IAM user.
		Following principle of least privilege access	In all node types, root user is not used (using special user "solana" instead).
	Application security	Security focused development practices	cdk-nag is being used with appropriate suppressions.
Cost optimization	Service selection	Use cost effective resources	1/ AMD-based instances are used for Consensus and RPC node to save the costs. Consider compiling Graviton-based binaries to improve costs for compute. 2/ Cost-effective EBS gp3 are preferred instead of io2. 3/ Solana nodes generate a substantial amount of outgoing data traffic, which deeds to be addressed with non-technical means like getting private agreements with AWS.
	Cost awareness	Estimate costs	Single RPC node with `r6a.8xlarge` EBS gp3 volumes about 2549 GB with On-Demand pricing will cost around US$1,596.43 per month in the US East (N. Virginia) region. More cost-optimal option with 3 year Compute Savings plan the cost goes down to $962.84 USD. Additionally, the data transfer costs can be about $1,356.80 USD per month for 15TB of outgoing traffic.
Reliability	Resiliency implementation	Withstand component failures	This solution uses AWS Application Load Balancer with RPC nodes for high availability. Newly provisioned Solana nodes triggered by Auto Scaling get up and running in about 30-50 minutes.
	Data backup	How is data backed up?	Considering blockchain data is replicated by nodes automatically and Solana nodes sync from start within an hour, we don't use any additional mechanisms to backup the data.
	Resource monitoring	How are workload resources monitored?	Resources are being monitored using Amazon CloudWatch dashboards. Amazon CloudWatch custom metrics are being pushed via CloudWatch Agent.
Performance efficiency	Compute selection	How is compute solution selected?	Compute solution is selected based on best price-performance, i.e. AWS AMD-based Amazon EC2 instances.
	Storage selection	How is storage solution selected?	Storage solution is selected based on best price-performance, i.e. gp3 Amazon EBS volumes with optimal IOPS and throughput.
	Architecture selection	How is the best performance architecture selected?	We used a combination of recommendations from the Solana community and our own testing.
Operational excellence	Workload health	How is health of workload determined?	Health of workload is determined via AWS Application Load Balancer Target Group Health Checks, on port 8899.
Sustainability	Hardware & services	Select most efficient hardware for your workload	The solution uses AMD-powered instances. There is a potential to use AWS Graviton-based Amazon EC2 instances which offer the best performance per watt of energy use in Amazon EC2.

Recommended Infrastructure

Usage pattern	Ideal configuration	Primary option on AWS	Data Transfer Estimates	Config reference
1/ Consensus node	32 vCPU, 256 GB RAM, Accounts volume: 1TB, 5K IOPS, 700 MB/s throughput, Data volume: 3TB, 10K IOPS, 700 MB/s throughput	r6a.8xlarge, Accounts volume: EBS gp3 1TB, 5K IOPS, 700 MB/s throughput, Data volume: EBS gp3 10K IOPS, 700 MB/s throughput	Proportional to the amount at stake. Between 200TB to 400TB/month	.env-sample-consensus
2/ Base RPC node (no secondary indexes)	32 vCPU, 256 GB RAM, Accounts volume: 1TB, 5K IOPS, 700 MB/s throughput, Data volume: 3TB, 12K IOPS, 700 MB/s throughput	r6a.8xlarge, Accounts volume: EBS gp3 1TB, 5K IOPS, 700 MB/s throughput Data volume: EBS gp3 12K IOPS, 700 MB/s throughput	150-200TB/month (no staking)	.env-sample-baserpc
3/ Extended RPC node (with all secondary indexes)	64 vCPU, 1 TB RAM, Accounts volume: 1TB, 7K IOPS, 700 MB/s throughput, Data volume: 3TB, 16K IOPS, 700 MB/s throughput	x2idn.16xlarge, Accounts: instance storage (ephemeral NVMe volumes) 1.9 TB, Data volume: 3TB, 12K IOPS, 700 MB/s throughput	150-200TB/month (no staking)	.env-sample-extendedrpc

Setup Instructions

Setup Cloud9

We will use AWS Cloud9 to execute the subsequent commands. Follow the instructions in Cloud9 Setup

Clone this repository and install dependencies

   git clone https://github.com/aws-samples/aws-blockchain-node-runners.git
   cd aws-blockchain-node-runners
   npm install

Deploy Single Node

Make sure you are in the root directory of the cloned repository
If you have deleted or don't have the default VPC, create default VPC

    aws ec2 create-default-vpc

NOTE: You may see the following error if the default VPC already exists: An error occurred (DefaultVpcAlreadyExists) when calling the CreateDefaultVpc operation: A Default VPC already exists for this account in this region.. That means you can just continue with the following steps.

Configure your setup

Create your own copy of .env file and edit it to update with your AWS Account ID and Region:

   # Make sure you are in aws-blockchain-node-runners/lib/solana
   cd lib/solana
   pwd
   cp ./sample-configs/.env-sample-baserpc .env
   nano .env

NOTE: Example configuration parameters are set in the local .env-sample file. You can find more examples inside sample-configs directory.

Deploy common components such as IAM role, and Amazon S3 bucket to store data snapshots

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana
   npx cdk deploy solana-common

Deploy Sync Node

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana
   npx cdk deploy solana-single-node --json --outputs-file single-node-deploy.json

After starting the node you need to wait for the initial synchronization process to finish. It may take about 30 minutes and you can use Amazon CloudWatch to track the progress. There is a script that publishes CloudWatch metrics every 5 minutes, where you can watch current block and slots behind metrics. When the node is fully synced the slots behind metric should go to 0. To see them:
- Navigate to CloudWatch service (make sure you are in the region you have specified for AWS_REGION)
- Open Dashboards and select solana-single-node from the list of dashboards.
Connect with the RPC API exposed by the node:

   INSTANCE_ID=$(cat single-node-deploy.json | jq -r '..|.nodeinstanceid? | select(. != null)')
   NODE_INTERNAL_IP=$(aws ec2 describe-instances --instance-ids $INSTANCE_ID --query 'Reservations[*].Instances[*].PrivateIpAddress' --output text)
    # We query token balance this account: https://solanabeach.io/address/9WzDXwBbmkg8ZTbNMqUxvQRAyrZzDsGYdLVL9zYtAWWM
    curl http://$NODE_INTERNAL_IP:8899 -X POST -H "Content-Type: application/json" \
    --data '{ "jsonrpc": "2.0", "id": 1, "method": "getBalance", "params": ["9WzDXwBbmkg8ZTbNMqUxvQRAyrZzDsGYdLVL9zYtAWWM"]}'

Deploy the HA Nodes

Configure and deploy multiple HA Nodes

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana
   npx cdk deploy solana-ha-nodes --json --outputs-file ha-nodes-deploy.json

Give the new RPC nodes about 30 minutes to initialize and then run the following query against the load balancer behind the RPC node created

    export RPC_ABL_URL=$(cat ha-nodes-deploy.json | jq -r '..|.ALBURL? | select(. != null)')
    echo $RPC_ABL_URL

    # We query token balance this account: https://solanabeach.io/address/9WzDXwBbmkg8ZTbNMqUxvQRAyrZzDsGYdLVL9zYtAWWM
    curl http://$RPC_ABL_URL:8899 -X POST -H "Content-Type: application/json" \
    --data '{ "jsonrpc": "2.0", "id": 1, "method": "getBalance", "params": ["9WzDXwBbmkg8ZTbNMqUxvQRAyrZzDsGYdLVL9zYtAWWM"]}'

The result should be like this (the actual balance might change):

   {"jsonrpc":"2.0","result":{"context":{"apiVersion":"1.16.15","slot":221433176},"value":12870473061872488},"id":1}

If the nodes are still starting and catching up with the chain, you will see the following response:

   <html>
   <head><title>503 Service Temporarily Unavailable</title></head>
   <body>
   <center><h1>503 Service Temporarily Unavailable</h1></center>
   </body>

NOTE: By default and for security reasons the load balancer is available only from within the default VPC in the region where it is deployed. It is not available from the Internet and is not open for external connections. Before opening it up please make sure you protect your RPC APIs.

Clearing up and undeploy everything

Undeploy HA Nodes, Single Nodes and Common stacks

   # Setting the AWS account id and region in case local .env file is lost
    export AWS_ACCOUNT_ID=<your_target_AWS_account_id>
    export AWS_REGION=<your_target_AWS_region>

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana

   # Undeploy HA Nodes
    cdk destroy solana-ha-nodes

    # Undeploy Single Node
    cdk destroy sync-single-node

    # Delete all common components like IAM role and Security Group
    cdk destroy solana-common

Follow steps to delete the Cloud9 instance in Cloud9 Setup

FAQ

How to check the logs of the clients running on my sync node?

Note: In this tutorial we chose not to use SSH and use Session Manager instead. That allows you to log all sessions in AWS CloudTrail to see who logged into the server and when. If you receive an error similar to SessionManagerPlugin is not found, install Session Manager plugin for AWS CLI

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana

   export INSTANCE_ID=$(cat single-node-deploy.json | jq -r '..|.node-instance-id? | select(. != null)')
   echo "INSTANCE_ID=" $INSTANCE_ID
   aws ssm start-session --target $INSTANCE_ID --region $AWS_REGION
   sudo su bcuser
   sudo journalctl -o cat -fu sol

How to check the logs from the EC2 user-data script?

   pwd
   # Make sure you are in aws-blockchain-node-runners/lib/solana

   export INSTANCE_ID=$(cat single-node-deploy.json | jq -r '..|.node-instance-id? | select(. != null)')
   echo "INSTANCE_ID=" $INSTANCE_ID
   aws ssm start-session --target $INSTANCE_ID --region $AWS_REGION
   sudo cat /var/log/cloud-init-output.log

How can I restart the Solana service?

   export INSTANCE_ID=$(cat single-node-deploy.json | jq -r '..|.node-instance-id? | select(. != null)')
   echo "INSTANCE_ID=" $INSTANCE_ID
   aws ssm start-session --target $INSTANCE_ID --region $AWS_REGION
   sudo systemctl status sol

How to upload a secret to AWS Secrets Manager?

    # Create key pair
    sudo ./solana-keygen new --no-passphrase -o /tmp/keypair.json
    SOLANA_ADDRESS=$(sudo ./solana-keygen pubkey /tmp/keypair.json)
    # Upload key pair to AWS Secrets Manager"
    export AWS_REGION=<your_region>
    sudo aws secretsmanager create-secret --name "solana/"$SOLANA_ADDRESS --description "Solana secret key pair" --secret-string file:///tmp/keypair.json --region $AWS_REGION
    #Delete key pair from the local file system
    rm -rf /tmp/keypair.json

How can I add swap space to the instance if my Solana node runs out of RAM during the initial sync?

There are two ways. Using the existing volume or using a new one. If your instance has Instance Store volume attached, it is better to keep your swap on it.
- Option 1: Dedicated Instance Store volume

   sudo mkswap /dev/nvme3n1
   sudo swapon /dev/nvme3n1
   # Check the memory space is updated
   free -g

Option 2: Existing volume (using Data directory as example):

   sudo mkdir /var/solana/data/swapfile
   sudo dd if=/dev/zero of=/var/solana/data/swapfile bs=1MiB count=250KiB
   sudo chmod 0600 /var/solana/data/swapfile
   sudo mkswap /var/solana/data/swapfile
   sudo swapon /var/solana/data/swapfile
   free -g
   sudo sysctl vm.swappiness=10

Upgrades

When nodes need to be upgraded or downgraded, use blue/green pattern to do it. This is not yet automated and contributions are welcome!

Sample AWS Blockchain Node Runner app for Solana Nodes

Overview of Deployment Architectures for Single and HA setups​

Single node setup​

HA setup​

Additional materials​

Setup Instructions​

Setup Cloud9​

Clone this repository and install dependencies​

Deploy Single Node​

Deploy the HA Nodes​

Clearing up and undeploy everything​

FAQ​

Upgrades​