web3-proxy/README.md
Bryan Stitt 2080739865
Quick cache (#70)
* quick cache and allocate less

* improve /status cache

* prepare to cache raw transaction hashes so we dont dos our backends

* simple benchmark for /health and /status

* mut not needed with atomics

* DRY all the status pages

* use u64 instead of bytes for subscriptions

* fix setting earliest_retry_at and improve logs

* Revert "use kanal instead of flume or tokio channels (#68)"

This reverts commit 510612d343fc51338a8a4282dcc229b50097835b.

* fix automatic retries

* put relaxed back

* convert error message time to seconds

* assert instead of debug_assert while we debug

* ns instead of seconds

* disable peak_latency for now

* null is the default

* cargo fmt

* comments

* remove request caching for now

* log on exit

* unit weigher for now

* make cache smaller. we need a weigher for prod. just debugging

* oops. we need async

* add todo

* no need for to_string on a RawValue
2023-05-13 11:13:02 -07:00

6.7 KiB

web3_proxy

Web3_proxy is a fast caching and load balancing proxy for web3 (Ethereum or similar) JsonRPC servers.

Under construction! This code is under active development. If you want to run this proxy youself, send me a message on Twitter and I can explain things that aren't documented yet. Most RPC methods are supported, but filters are coming soon. And of course, more tests are always needed.

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 (llamanodes, alchemy, moralis, rivet, your own node, or one of many other providers). If multiple servers are in sync, they are prioritized by active_requests and request latency. 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 server 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.

Quick development

  1. brew install librdkafka or sudo apt-get install librdkafka-dev
  2. Run docker-compose up -d to start the database and caches. See docker-compose.yml for details.
  3. Copy ./config/example.toml to ./config/development.toml and change settings to match your setup.
  4. Run cargo commands:
$ cargo run --release -- --help
   Compiling web3_proxy v0.1.0 (/home/bryan/src/web3_proxy/web3_proxy)
    Finished release [optimized + debuginfo] target(s) in 17.69s
     Running `target/release/web3_proxy --help`
Usage: web3_proxy [--port <port>] [--workers <workers>] [--config <config>]

web3_proxy is a fast caching and load balancing proxy for web3 (Ethereum or similar) JsonRPC servers.

Options:
  --port            what port the proxy should listen on
  --workers         number of worker threads
  --config          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/development.toml which uses the database and cache running under docker and proxies to a bunch of public nodes:

cargo run --release -- proxyd

Common commands

Create a user:

cargo run -- --db-url "$YOUR_DB_URL" create_user --address "$USER_ADDRESS_0x"

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

Check that the websocket is working:

$ websocat ws://127.0.0.1:8544

{"jsonrpc": "2.0", "id": 1, "method": "eth_subscribe", "params": ["newHeads"]}

{"jsonrpc": "2.0", "id": 2, "method": "eth_subscribe", "params": ["newPendingTransactions"]}

{"jsonrpc": "2.0", "id": 3, "method": "eth_subscribe", "params": ["newPendingFullTransactions"]}

{"jsonrpc": "2.0", "id": 4, "method": "eth_subscribe", "params": ["newPendingRawTransactions"]}

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

Compare 3 RPCs:

web3_proxy_cli health_compass https://eth.llamarpc.com https://eth-ski.llamarpc.com https://rpc.ankr.com/eth

Run migrations

Generally it is simplest to just run the app to run migrations. It runs migrations on start.

But if you want to run them manually (generally only useful in development):

cd migration
cargo run up

Create a user:

web3_proxy_cli --config ... create_user --address 0x0000000000000000000000000000000000000000 --email infra@llamanodes.com --description "..."

Give a user unlimited requests per second:

Copy the ULID key (or UUID key) out of the above command, and put it into the following command.

web3_proxy_cli --config ... change_user_tier_by_key "$RPC_ULID_KEY_FROM_PREV_COMMAND" "Unlimited"

Health compass

Health check 3 servers and error if the first one doesn't match the others.

web3_proxy_cli health_compass https://eth.llamarpc.com/ https://rpc.ankr.com/eth https://cloudflare-eth.com

Adding new database tables

cargo install sea-orm-cli
  1. (optional) drop the current dev db
  2. sea-orm-cli migrate
  3. sea-orm-cli generate entity -u mysql://root:dev_web3_proxy@127.0.0.1:13306/dev_web3_proxy -o entities/src --with-serde both
  4. After running the above, you will need to manually fix some columns: Vec<u8> -> sea_orm::prelude::Uuid and i8 -> bool. Related: https://github.com/SeaQL/sea-query/issues/375 https://github.com/SeaQL/sea-orm/issues/924

Flame Graphs

Flame graphs make a developer's join of finding slow code painless:

$ cat /proc/sys/kernel/kptr_restrict
1
$ echo 0 | sudo tee /proc/sys/kernel/kptr_restrict
0
$ cat /proc/sys/kernel/perf_event_paranoid
4
$ echo -1 | sudo tee /proc/sys/kernel/perf_event_paranoid
-1
$ CARGO_PROFILE_RELEASE_DEBUG=true cargo flamegraph --bin web3_proxy_cli --no-inline -- proxyd

Be sure to use --no-inline or perf will be VERY slow

GDB

Developers can run the proxy under gdb for advanced debugging:

cargo build --release && RUST_LOG=info,web3_proxy=debug,ethers_providers::rpc=off rust-gdb --args target/debug/web3_proxy --listen-port 7503 --rpc-config-path ./config/production-eth.toml

TODO: also enable debug symbols in the release build by modifying the root Cargo.toml

Load Testing

Test the proxy:

wrk -t12 -c400 -d30s --latency http://127.0.0.1:8544/health
wrk -t12 -c400 -d30s --latency http://127.0.0.1:8544/status
wrk -s ./wrk/getBlockNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8544/u/$API_KEY
wrk -s ./wrk/getLatestBlockByNumber.lua -t12 -c400 -d30s --latency http://127.0.0.1:8544/u/$API_KEY

Test geth (assuming it is on 8545):

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

Test erigon (assuming it is on 8945):

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

Run ethspam and versus for a more realistic load test:

ethspam --rpc http://127.0.0.1:8544 | versus --concurrency=100 --stop-after=10000 http://127.0.0.1:8544

ethspam --rpc http://127.0.0.1:8544/u/$API_KEY | versus --concurrency=100 --stop-after=10000 http://127.0.0.1:8544/u/$API_KEY