index what matters
Filter by program owner and keep the useful part of Solana state instead of carrying every account in the read path.
[ status: live ] [ chain: solana ] [ mode: indexed account rpc ]
fast account reads for solana, served like infrastructure.
Cloudbreak keeps selected Solana account state in PostgreSQL, ingests live updates from Yellowstone gRPC, bootstraps from snapshots, and exposes a focused JSON-RPC surface for account-heavy workloads.
// about
not another full validator rpc.
Cloudbreak separates high-volume account queries from the validator path. It acts as a dedicated account read layer that indexes the state you care about, then serves it through familiar Solana-style endpoints.
bootstrap fast
Load full and incremental snapshots, then catch up using Yellowstone account and slot streams.
optimize live traffic
The query tracker can observe frequent filters and create PostgreSQL indexes for repeated workloads.
serve focused rpc
Built around methods like getProgramAccounts, token account queries, getMultipleAccounts, and getBalance.
// pipeline
source → ingest → state → serve
The data path is intentionally direct. Live chain data and snapshots feed the indexer, PostgreSQL holds indexed account state, and the API server exposes it over HTTP JSON-RPC.
yellowstone gRPC
Account updates, slot notifications, and finalization state.
snapshots
Full and incremental snapshots provide the initial dataset.
indexer
Filters state, handles slot continuity, and applies updates to the store.
postgresql
Accounts, slots, and query-path indexes live here.
api server
HTTP JSON-RPC for bots, explorers, wallets, and internal infrastructure.
// methods
familiar endpoints, different backend.
Cloudbreak is intentionally narrow. It focuses on account-state methods where indexing and database-backed reads help the most.
selected_method
getProgramAccounts
Returns accounts owned by a program with memcmp, dataSize, and dataSlice support. Responses can be streamed as chunked JSON-RPC frames.
POST / HTTP/1.1
content-type: application/json
{"jsonrpc":"2.0","id":1,"method":"getProgramAccounts"}// benchmarks
included repo numbers, shown as engineering data.
The repository includes benchmark material dated June 29, 2026. These numbers reflect the documented workload and test setup. They should be read as project data, not marketing promises.
Cloudbreak p50 in the low-RPS 0–1 KB response comparison shown in the repo benchmark notes.
Effective requests per second reported for the EPYC Gen3 benchmark case.
Effective requests per second reported for the EPYC Gen4 benchmark case.
// quick-start
spin it up like infrastructure.
Cloudbreak is a Rust workspace that expects a running PostgreSQL database and access to snapshots plus a Yellowstone gRPC source.
rust
postgresql
cargo
yellowstone grpc
solana snapshots
clone the repo
git clone https://github.com/solana-rpc/cloudbreak.git
configure environment
cp .env.example .env
Point Cloudbreak to PostgreSQL, snapshot files, and a Yellowstone gRPC source.
run ingest + api
cargo run --bin cloudbreak
After bootstrap, the API listens for JSON-RPC requests and serves the indexed state.
// deploy