Submitting a transaction to be executed on the solana blockchain, requires the client to identify the next few leaders based on the leader schedule, look up their peering information in gossip and connect to them via the quic protocol. In order to simplify the process so it can be triggered from a web browser, most applications run full validators that forward the transactions according to the protocol on behalf of the web browser. Running full solana validators is incredibly resource intensive (>256GB RAM), the goal of this project would be to create a specialized micro-service that allows to deploy this logic quickly and allows horizontal scalability with commodity vms. Optionally the Lite RCP micro-services can be configured to send the transactions to a complementary QUIC forward proxy instead of the solana tpu (details).

1. Subscribe to new blocks using websockets (deprecated)
2. Polling blocks over RPC.(Current)
3. Subscribe blocks over gRPC. (Current)
4. Listening to gossip protocol. (Future roadmap)

run using

$ cargo run --release

to know about command line options

$ cargo run --release -- --help

Make sure both solana-validator and lite-rpc is running

test

$ cargo test

bench

$ cd bench
$ RUST_LOG=info cargo run -- --help

Find a new file named metrics.csv in the project root.

Thank you for providing the default values. Here's the updated table with the default values for the environment variables based on the additional information:

lite-rpc implements an optional postgres service that can write to postgres database tables as defined in ./migrations. This can be enabled by either setting the environment variable PG_ENABLED to true or by passing the -p option when launching the executable. If postgres is enabled then the optional environment variables shown above must be set.

Various Prometheus metrics are exposed on localhost:9091/metrics which can be used to monitor the health of the application in production.

While lite-rpc can be deployed on any cloud infrastructure, it has been tested extensively on https://fly.io. An example configuration has been provided in fly.toml. We recommend a dedicated-cpu-2x VM with at least 4GB RAM.

The app listens by default on ports 8890 and 8891 for HTTP and Websockets respectively. Since only a subset of RPC methods are implemented, we recommend serving unimplemented methods from a full RPC node using a reverse proxy such as HAProxy or Kong. Alternatively, you can connect directly to lite-rpc using a web3.js Connection object that is only used for sending and confirming transactions.

Troubleshooting: if you encounter issues with QUIC sendmsg check this - you might need to explicitly disable GSO (Generic Segmenatin Offload) see DISABLE_GSO=true

fly apps create my-lite-rpc
fly secrets set -a my-lite-rpc RPC_URL=... WS_URL=...   # See above table for env options
fly scale vm dedicated-cpu-2x --memory 4096 -a my-lite-rpc
fly deploy -c cd/lite-rpc.toml -a my-lite-rpc --remote-only # To just launch lite-rpc
fly deploy -c cd/lite-rpc.toml -a my-lite-rpc --remote-only # To launch lite-rpc with proxy mode

Copyright (c) 2022 Blockworks Foundation

Licensed under the AGPL-3.0 license