🧪 Performance Testing Blockchains using OctoPerf & Grafana

Objective: Measure performance of Ethereum node(s) under load by simulating JSON-RPC API traffic using OctoPerf and track system metrics via Grafana and InfluxDB.

🔧 PART 1: Environment Setup

1. Set Up Ethereum Node(s) in Docker

We'll use Geth for this example:

‍

docker run -d --name geth-node \
 -p 8545:8545 \
 -v $HOME/.ethereum:/root/.ethereum \
 ethereum/client-go \
 --http --http.addr "0.0.0.0" --http.api "eth,net,web3" --http.corsdomain "*" --syncmode "fast"


💡 Use multiple nodes and expose JSON-RPC interfaces if testing cluster performance.

2. Set Up InfluxDB for Metrics Storage

‍

‍

docker run -d --name=influxdb \
 -p 8086:8086 \
 -v influxdb:/var/lib/influxdb \
 influxdb:1.8


Create a database for metrics:

docker exec -it influxdb influx
> CREATE DATABASE eth_metrics;
> exit


3. Set Up Grafana

‍

docker run -d --name=grafana \
 -p 3000:3000 \
 --link influxdb \
 grafana/grafana


• Access Grafana: http://localhost:3000 (default login: admin/admin)
• Add InfluxDB as a data source (URL: http://influxdb:8086, Database: eth_metrics)

📈 PART 2: Enable System Metrics Collection

Install Telegraf or Docker Stats Exporter to collect metrics and push to InfluxDB:

Option A: Telegraf (Host Monitoring)

docker run -d --name=telegraf \
 --net=host \
 -v /proc:/host/proc:ro \
 -v /sys:/host/sys:ro \
 -v /etc:/host/etc:ro \
 -e HOST_PROC=/host/proc \
 -e HOST_SYS=/host/sys \
 -e HOST_ETC=/host/etc \
 telegraf \
 --config-directory /etc/telegraf/telegraf.d


Configure telegraf.conf to send to InfluxDB:

toml

[[outputs.influxdb]]
 urls = ["http://localhost:8086"]
 database = "eth_metrics"


🚀 PART 3: Configure OctoPerf for Load Testing

1. Create an OctoPerf Account

• Go to https://app.octoperf.com
• Create a free or paid account.

2. Build the Test Plan

1. Create a New Project
2. Create a Virtual User
   • Choose "HTTP Request"
   • Set method to POST
   • URL: http://<host>:8545
   • Body:
   jsonCopyEdit
   {  "jsonrpc":"2.0",  "method":"eth_blockNumber",  "params":[],  "id":1}
   • Header: Content-Type: application/json
3. Add Validation (Optional)
   • Add assertions to check response contains "result" or valid block number.
4. Add Think Time (Optional)
   • Simulate realistic load using pacing or think-time.
5. Parameterization (Optional)
   • Load addresses or methods dynamically for varied testing.

3. Run the Test

• Choose number of users (threads), ramp-up, and duration.
• Deploy from a cloud location or from your local agent (Docker agent available).
• Start test and collect metrics: latency, throughput, errors, etc.

📊 PART 4: Real-Time Monitoring with Grafana

Suggested Dashboards

• Node Health Dashboard: CPU, memory, disk, network I/O
• RPC Performance Dashboard:
  • InfluxDB queries for HTTP request metrics (via Telegraf or custom exporters)
• Custom RPC Metrics (Advanced):
  • Use Prometheus exporters (e.g., eth-exporter or custom scripts)

Example InfluxDB query (for Grafana panel):

sql

SELECT mean("usage_cpu") FROM "cpu" WHERE $timeFilter GROUP BY time($__interval), "host"


🧼 PART 5: Post-Test Analysis

From OctoPerf

• Analyze:
  • Response time percentiles
  • Error rates
  • Throughput trends
• Export reports as PDF/CSV.

From Grafana

• Review system utilization trends correlated with test phases
• Check for:
  • CPU saturation
  • Memory leaks (JVM for Besu)
  • Network congestion
  • Disk I/O spikes (if syncing)

🛠️ Troubleshooting Tips

SymptomCauseFix500 Errors in OctoPerfRate-limited RPC callsIncrease node resources or rate-limit trafficGrafana shows no dataTelegraf misconfiguredCheck output section of telegraf.confLow throughputNode not syncedEnsure Geth/Besu is fully syncedTest fails quicklyDocker bridge issuesCheck Docker network mode and firewall rules

🧪 Optional Enhancements

• Add Prometheus & eth-exporter for Ethereum-specific metrics.
• Deploy a Geth private network for full control of blockchain state.
• Use k6 or JMeter for additional comparison with OctoPerf.
• Automate test cycles via CI/CD pipelines (e.g., Jenkins or GitHub Actions).

‍