web3-proxy

Web3-proxy is a fast caching and load balancing proxy for web3 (Ethereum or similar) JsonRPC servers.

Signed transactions (eth_sendRawTransaction) are sent in parallel to the configured private RPCs (eden, ethermine, flashbots, etc.).

All other requests are sent to an RPC server on the latest block (alchemy, moralis, rivet, your own node, or one of many other providers). If multiple servers are in sync, they are prioritized by active_requests/soft_limit. Note that this means that the fastest server is most likely to serve requests and slow servers are unlikely to ever get any requests.

Each servers has different limits to configure. The soft_limit is the number of parallel active requests where a server starts to slow down. The hard_limit is where a server starts giving rate limits or other errors.

cargo run --release -- --help

   Compiling web3-proxy v0.1.0 (/home/bryan/src/web3-proxy/web3-proxy)
    Finished release [optimized] target(s) in 9.45s
     Running `target/release/web3-proxy --help`
Usage: web3-proxy [--listen-port <listen-port>] [--rpc-config-path <rpc-config-path>]

Web3-proxy is a fast caching and load balancing proxy for web3 (Ethereum or similar) JsonRPC servers.

Options:
  --listen-port     what port the proxy should listen on
  --rpc-config-path path to a toml of rpc servers
  --help            display usage information

Start the server with the defaults (listen on http://localhost:8544 and use ./config/example.toml which proxies to a local websocket on 8546 and ankr's public ETH node):

cargo run --release

Check that the proxy is working:

curl -X POST -H "Content-Type: application/json" --data '{"jsonrpc":"2.0","method":"web3_clientVersion","id":1}' 127.0.0.1:8544/eth

You can copy config/example.toml to config/production-$CHAINNAME.toml and then run docker-compose up --build -d start a proxies for many chains.

Flame Graphs

Flame graphs make finding slow code painless:

$ cat /proc/sys/kernel/kptr_restrict
1
$ echo 0 |sudo tee /proc/sys/kernel/kptr_restrict
0
$ CARGO_PROFILE_RELEASE_DEBUG=true cargo flamegraph

Load Testing

Test the proxy:

wrk -s ./data/wrk/getBlockNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8544
wrk -s ./data/wrk/getLatestBlockByNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8544

Test geth:

wrk -s ./data/wrk/getBlockNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8545
wrk -s ./data/wrk/getLatestBlockByNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8545

Test erigon:

wrk -s ./data/wrk/getBlockNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8945
wrk -s ./data/wrk/getLatestBlockByNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8945

Note: Testing with getLatestBlockByNumber.lua is not great because the latest block changes and so one run is likely to be very different than another.

Todo

• after connecting to a server, check that it gives the expected chainId
• if the fastest server has hit rate limits, we won't be able to serve any traffic until another server is synced.
• proper logging with useful instrumentation
• think more about how multiple rpc tiers should work
• if a request gets a socket timeout, try on another server
  • maybe always try at least two servers in parallel? and then return the first? or only if the first one doesn't respond very quickly?
• incoming rate limiting (by ip or by api key or what?)
• measure latency to nodes?
• one proxy for mulitple chains?
• zero downtime deploys
• are we using Acquire/Release/AcqRel properly? or do we need other modes?
• simple proxy
• better locking. when lots of requests come in, we seem to be in the way of block updates
• load balance between multiple RPC servers
• support more than just ETH
• option to disable private rpc and send everything to primary
• health check nodes by block height
• Dockerfile
• docker-compose.yml