Atlas Releases

What is the Atlas RPC Node

Atlas RPC nodes are bridges between end users and our Atlas sequencer.

It provides a Solana-compatible JSON-RPC and websocket endpoint.

All transactions sent to the RPC node will be forwarded to our sequencer.

Once the sequencer processed those transactions, they would be sent back to the RPC node by a message bus (Redis for now), the RPC node would replay the transaction to make sure that the sequencer is not malicious.

Machine spec:

At least 32 cores, 64G memory, and 100G disk linux/amd64 or linux/arm64

We support machines with lower specs, but we recommend that you run the below commands with the --low-cpu flag.

Upstream connections / External Dependencies:

Redis(Message Bus): redis://redis-testnet.atlas.xyz:6379/

Postgres(Serve historical data): postgresql://public_access:[email protected]:5432/svm_node

Sequencer: http://testnet.atlas.xyz:3002 (API Key required)

Start the server from binary:

atlas-replay-node --mode=rpc \
    --log-level=info \
    --redis-url='redis://redis-testnet.atlas.xyz:6379/' \
    --server-url='https://testnet.atlas.xyz:3002/' \
    --postgres-url='postgresql://public_access:[email protected]:5432/svm_node' \
    --api-key=<YOUR API KEY> \
    --num-async-threads=2

Low CPU mode

This will use up to 4 CPU cores

atlas-replay-node --mode=rpc \
    --log-level=info \
    --redis-url='redis://redis-testnet.atlas.xyz:6379/' \
    --server-url='https://testnet.atlas.xyz:3002/' \
    --postgres-url='postgresql://public_access:[email protected]:5432/svm_node' \
    --api-key=<YOUR API KEY> \
    --num-async-threads=1 \
    --low-cpu

To save logs elsewhere, pass the argument --log-dir. The log files on disk will be rotated daily.

Start from docker

docker run -d \
        --name atlas-rpc \
        -p 8899:8899 -p 8900:8900 \
        ghcr.io/ellipsis-labs/atlas-replay-node:v0.0.1 \
        --mode=rpc \
        --log-level=info \
        --redis-url='redis://redis-testnet.atlas.xyz:6379/' \
        --server-url='https://testnet.atlas.xyz:3002/' \
        --postgres-url='postgresql://public_access:[email protected]:5432/svm_node' \
        --api-key=<YOUR API KEY> \
        --num-async-threads=1

Low CPU mode

This will use up to 4 CPU cores

docker run -d \
        --name atlas-rpc \
        -p 8899:8899 -p 8900:8900 \
        ghcr.io/ellipsis-labs/atlas-replay-node:v0.0.1 \
        --mode=rpc \
        --log-level=info \
        --redis-url='redis://redis-testnet.atlas.xyz:6379/' \
        --server-url='https://testnet.atlas.xyz:3002/' \
        --postgres-url='postgresql://public_access:[email protected]:5432/svm_node' \
        --api-key=<YOUR API KEY> \
        --num-async-threads=1 \
        --low-cpu

Kubernetes template

apiVersion: apps/v1
kind: Deployment
metadata:
  name: atlas-rpc
  labels:
    app: atlas-rpc
spec:
  replicas: 3
  selector:
    matchLabels:
      app: atlas-rpc
  template:
    metadata:
      labels:
        app: atlas-rpc
    spec:
      containers:
      - name: atlas-rpc
        image: ghcr.io/ellipsis-labs/atlas-replay-node:v0.0.1
        ports:
        - containerPort: 8899
        - containerPort: 8900
        env:
        args:
        - --mode=rpc
        - --log-level=info
        - --redis-url=redis://redis-testnet.atlas.xyz:6379/
        - --server-url=https://testnet.atlas.xyz:3002/
        - --api-key=<YOUR API KEY>
        - --postgres-url=postgresql://public_access:[email protected]:5432/svm_node
        - --num-async-threads=2
        resources:
          requests:
            cpu: 32
            memory: 64Gi
          limits:
            cpu: 32
            memory: 64Gi
        livenessProbe:
          httpGet:
            path: /liveness
            port: 8899
          initialDelaySeconds: 5
          periodSeconds: 3
        readinessProbe:
          httpGet:
            path: /readiness
            port: 8899
          initialDelaySeconds: 5
          periodSeconds: 1

Misc

• Historical state storage

  Right now, the historical state is saved in our Postgres database.

  In the future, once we make our transaction ledger public, anyone can replay the ledger, rebuild the historical state, and store it in a private database.

• Re: external dependencies, some may change - We’re planning to migrate our Redis to Kafka soon, and the historical storage layer hasn’t been finalized yet.