Skip to content

shalzz/ethereum-worker

Repository files navigation

Ethereum worker

Run your own Ethereum node with CDN caching on a Raspberry Pi or the cloud

About

A scaling solution for Ethereum nodes by caching responses using Clouldflares technologies like Argo tunnels and Cloudflare workers.

This is very similar to the official Cloudflare Ethereum Gateway service provided by Cloudflare here but allows you to run and configure your own geth or parity node and put it behind a global edge computing network with robust caching facilities.

Philosophy

With the immutable property of blockchains and a block based model it is easier to aggressively cache a response without worrying about invalidating a cache on updates/writes within a block rather than indexing individual transactions and blocks to store in a database which are unlikely to change save for the latest few blocks (80-100 for ethereum) to account for a chain re-org.

Which is why we, among others like Cloudflares Ethereum Gateway service, have decided to go with this approach.

Demo

We have a demo worker running that you can use to quickly try out how it works at: https://ethereum-worker.8bitlabs.workers.dev

For example, query the web3_clientVersion using curl

$ curl -d '{"method":"web3_clientVersion","params":[],"id":1,"jsonrpc":"2.0"}' https://ethereum-worker.8bitlabs.workers.dev

Try it for free

Cloudflare provides free modes of both the services we will be using.

Install

To deploy a Cloudflare worker we need to install the wrangler CLI tool from Cloudflare. Follow the instructions here to install it on your machine.

Once you have wrangler installed, create a Cloudflare account if you haven't already and configure wrangler with your account using

wrangler config

You'll need to provide your email associated with your account and your Global API Key.

For help on how to create an API key see this support article. It is possible to use a scope limited API key for this but it is not well documented and you will have to do some trial and error to get it working.

Before we can publish our worker to enable us to serve cached responses for the same kinds of requests we need to specify the ORIGIN_URL which points to the server where we have our Ethereum node running.

The ORIGIN_URL can be the IP address of our server with the port specified listening for RPC requests or it can be a load balancer domain name. But with Cloudflare workers an Argo tunnel is recommended since they work well together and are optimised for each other with the added benefit of Argo tunnels being cheaper compare to the prices usually charged by major cloud providers for a load balancer.

If you don't already have a Ethereum client running see below for helm charts that can help you get one running. If you have a client running and just want to setup an Argo tunnel see official docs here.

To change the ORIGIN_URL edit the wrangler.toml file and specify the URL as an environment variable.

vars = { ORIGIN_URL = "https://url-pointing-to-my-ethereum-node.com"}

Now we are ready to publish. Make sure we have workers_dev = true set so that we only use the free workers tier plan.

Write Specific Endpoint

Optionally you can specify a separate endpoint to use for write operations, for example the flashbots RPC endpoint.

To do so, specify a WRITE_URL var in your wrangler.toml file

vars = { ORIGIN_URL = "https://url-pointing-to-my-ethereum-node.com", WRITE_URL = "https://rpc.flashbots.net" }

Publish

To publish all we need to do now it run the command

wrangler publish

And if everything went well you should now see a workers.dev URL that you can use to query your web3 JSON-RPC requests utilizing the CDN and edge computing capabilities of the Cloudflare network!

Helm Charts

If you are looking to quickly deploy an Ethereum client with Argo tunnels setup, we provide Helm Charts for deploying the Ethereum clients geth and parity to a Kubernetes clusters with a Argo tunnel running as a side car model.

See the helm-chart folder and chart specific README.md for more details.