RPC Best Practices

A condensed summary of the 25 most important RPC practices drawn from every page in this section.

Never ship production on public RPC: Use a paid provider or self-hosted node with SLA (RPC Providers).
Store keys in environment variables: Never commit API keys; proxy sensitive reads through your backend.
Pair HTTP and WSS URLs: Subscriptions need wss:// from the same vendor dashboard (WebSocket Subscriptions).
Set commitment explicitly: Default to confirmed for UX; finalized for treasury (RPC Basics).
Null-check getAccountInfo: value: null means missing account - do not decode blindly (Core RPC Methods).
Prefer getMultipleAccounts: Batch known pubkeys instead of N serial getAccountInfo calls.
Never run unfiltered GPA on mainnet: Always memcmp/dataSize filter or use an indexer (getProgramAccounts & Filters).
Chunk large batches: 50-100 accounts per getMultipleAccounts request (Pagination & Efficiency).
Paginate signatures server-side: getSignaturesForAddress cursoring belongs in your API, not mobile clients for whales.
Abort WebSocket subscriptions: Clean up on unmount to avoid connection leaks.
Use signatureNotifications for confirms: Prefer subscriptions over 200ms polling loops.
Simulate before mainnet sends: Read unitsConsumed and logs (Priority Fee & Simulation APIs).
Refresh priority fees per batch: getRecentPrioritizationFees stale samples underprice during congestion.
Add CU headroom: Set compute unit limit to 110% of simulated consumption.
Use DAS for wallet NFTs: getAssetsByOwner beats manual metadata GPA (The DAS API).
Keep DAS and RPC URLs documented: They may differ on provider dashboards.
Implement RPC failover: Secondary URL on 429/503 with exponential backoff.
Rate-limit your own clients: Protect provider quota from retry storms.
Align CLI and app RPC: solana config get should match RPC_URL when debugging.
Cache immutable data only: Short TTL for mint metadata; never cache post-swap balances without invalidation.
Monitor slot lag: getSlot RTT and stall detection alert unhealthy endpoints.
Avoid heavy getTransaction in browsers: Proxy through indexer for history views.
Use jsonParsed selectively: Fall back to base64 + IDL decode when parser unsupported.
Plan indexer migration early: If GPA or history scans exceed thousands of rows, start Geyser pipeline (Indexing Basics).
Pin @solana/kit 7.0.0: Typed RPC reduces param bugs vs raw fetch.

FAQs

What is the top RPC mistake?

Unfiltered mainnet getProgramAccounts in production - timeouts, cost, and bans.

Public RPC ever OK?

Local prototypes and rare CLI use only - not user-facing mainnet traffic.

How do RPC practices tie to tx landing?

Simulation and priority fee RPCs feed compute budget decisions - see tx optimization section.

RPC Basics - section entry
RPC Providers - vendor choice
Pagination & Efficiency - scale patterns
Indexing Best Practices - beyond RPC

Stack versions: This page was written for Agave 4.1.1, Solana CLI 3.0.10, Anchor 0.32.1, anchor-lang 0.32.1, Rust 1.91.1, @solana/kit 7.0.0, Surfpool 0.12.0, and LiteSVM 0.6.x.

RPC Best Practices

A condensed summary of the 25 most important RPC practices drawn from every page in this section.

Never ship production on public RPC: Use a paid provider or self-hosted node with SLA (RPC Providers).

Store keys in environment variables: Never commit API keys; proxy sensitive reads through your backend.

Pair HTTP and WSS URLs: Subscriptions need wss:// from the same vendor dashboard (WebSocket Subscriptions).

Set commitment explicitly: Default to confirmed for UX; finalized for treasury (RPC Basics).

Null-check getAccountInfo: value: null means missing account - do not decode blindly (Core RPC Methods).

Prefer getMultipleAccounts: Batch known pubkeys instead of N serial getAccountInfo calls.

Never run unfiltered GPA on mainnet: Always memcmp/dataSize filter or use an indexer (getProgramAccounts & Filters).

Chunk large batches: 50-100 accounts per getMultipleAccounts request (Pagination & Efficiency).

Paginate signatures server-side: getSignaturesForAddress cursoring belongs in your API, not mobile clients for whales.

Abort WebSocket subscriptions: Clean up on unmount to avoid connection leaks.

Use signatureNotifications for confirms: Prefer subscriptions over 200ms polling loops.

Simulate before mainnet sends: Read unitsConsumed and logs (Priority Fee & Simulation APIs).

Refresh priority fees per batch: getRecentPrioritizationFees stale samples underprice during congestion.

Add CU headroom: Set compute unit limit to 110% of simulated consumption.

Use DAS for wallet NFTs: getAssetsByOwner beats manual metadata GPA (The DAS API).

Keep DAS and RPC URLs documented: They may differ on provider dashboards.

Implement RPC failover: Secondary URL on 429/503 with exponential backoff.

Rate-limit your own clients: Protect provider quota from retry storms.

Align CLI and app RPC: solana config get should match RPC_URL when debugging.

Cache immutable data only: Short TTL for mint metadata; never cache post-swap balances without invalidation.

Monitor slot lag: getSlot RTT and stall detection alert unhealthy endpoints.

Avoid heavy getTransaction in browsers: Proxy through indexer for history views.

Use jsonParsed selectively: Fall back to base64 + IDL decode when parser unsupported.

Plan indexer migration early: If GPA or history scans exceed thousands of rows, start Geyser pipeline (Indexing Basics).

Pin @solana/kit 7.0.0: Typed RPC reduces param bugs vs raw fetch.

FAQs

What is the top RPC mistake?

Unfiltered mainnet getProgramAccounts in production - timeouts, cost, and bans.

Public RPC ever OK?

Local prototypes and rare CLI use only - not user-facing mainnet traffic.

How do RPC practices tie to tx landing?

Simulation and priority fee RPCs feed compute budget decisions - see tx optimization section.

Stack versions: This page was written for Agave 4.1.1, Solana CLI 3.0.10, Anchor 0.32.1, anchor-lang 0.32.1, Rust 1.91.1, @solana/kit 7.0.0, Surfpool 0.12.0, and LiteSVM 0.6.x.

RPC Best Practices

FAQs

Related

Search Documentation

RPC Best Practices

FAQs

Related