Map your Google Cloud workloadShared RPC or managed node
Verify your new endpointsHTTPS, WebSocket and archive reads
Switch production trafficCanary, monitoring and fallback
Choose your migration path
Google Cloud lists December 15, 2026 as the shutdown date for both services. Existing Google endpoints stop serving dependent workloads at shutdown. Quicknode is Google Cloud’s recommended migration partner.
Create an authenticated OnFinality endpoint for the same network. Replace your URL and update authentication in each client. Compare the methods your application uses before moving production traffic.
Map the capabilities of your full or archive node. Check historical state, trace methods and tracer options. Include separate Beacon or consensus endpoints, client settings, and isolation requirements in your migration plan.
Keep each workload on the same network when replacing a provider. Use the HTTPS and WSS URLs from your OnFinality dashboard. For legacy testnets such as Goerli or Holesky, choose a maintained testnet and redeploy test contracts and fixtures.
Copy the OnFinality URLs from your dashboard and remove Google-specific authentication, such as X-goog-api-key, from the new client. Keep the old credentials on the old route while you validate.
Ethereum and Polygon. Re-run eth_call and eth_getLogs queries with the same parameters. Test filters and subscriptions over WSS, and confirm access to debug_traceTransaction and the tracer options you use.
Solana. Compare getGenesisHash and common reads with the same commitment level. Run logsSubscribe or your app’s other subscription methods on the WebSocket endpoint.
Archive and custom nodes. Test the oldest state you read, plus any consensus API or client-specific namespace separately.
Before cutover, run this read-only check with Node.js 22 or later. It checks the chain ID and compares the block hash at a height both endpoints report as finalized. A failed check exits with a nonzero status.
Save the configuration as .env.migration beside the downloaded script and keep it out of version control. Use the full endpoint URLs, including query credentials where required. Then run:
Set ARCHIVE_ADDRESS and ARCHIVE_BLOCK to compare a historical balance. Use a hexadecimal block number at or below the common finalized height.
Set GOOGLE_WSS_URL and ONFINALITY_WSS_URL to confirm that each endpoint delivers an eth_subscribe newHeads event. The script uses URL-based WSS credentials. If your old socket requires a custom header, verify it separately with your application client.
Set EXPECTED_CHAIN_ID=11155111 for Sepolia. The default is Ethereum Mainnet (1).
If your Google endpoint is offline, compare OnFinality against another trusted endpoint on the same chain. Re-run your app’s checks for logs, traces, writes and reconnect behavior before switching traffic.
Production cutover checklist
Set acceptance and rollback thresholds before the first cutover. Treat HTTPS requests and persistent WebSocket connections as separate traffic paths.
Inventory the production workload
List networks, HTTPS and WSS clients, authentication, peak request rate, required methods and node settings. Include jobs, indexers, monitoring and fallback URLs.
Create and test the replacement
Use a separate OnFinality key for staging. Run the parity script, reproduce your busiest and most expensive reads, and rehearse reconnecting and backfilling subscriptions from a checkpoint.
Route a small share of traffic first
Keep both endpoints behind a configuration switch or weighted router. Start with read-only traffic and increase it as error rate, 429s, p95 latency, head lag and subscription gaps stay within your thresholds.
Switch writes and subscriptions
Keep one active writer. Before retrying a submission, check whether it was already accepted. Resume subscriptions from the last processed block or slot, and reconcile duplicates and gaps.
Rehearse rollback
Assign a rollback owner and set thresholds. While Google resources remain available, keep a tested previous route. Near or after shutdown, use an independently verified fallback on the same network.
Retire the old resources
After the observation window, remove Google URLs from clients and fallback configurations. Follow Google’s instructions to delete legacy Blockchain Node Engine nodes. Google removes Blockchain RPC endpoints at shutdown.
Migration FAQ
When do Google Cloud Blockchain Node Engine and Blockchain RPC shut down?
Google Cloud lists December 15, 2026 as the shutdown date. Remaining nodes and endpoints are scheduled for deletion.
Can I use OnFinality instead of Quicknode for this migration?
Yes. Quicknode is Google Cloud’s recommended partner. To evaluate OnFinality, create an endpoint and validate the methods and access required by your workload.
Standard JSON-RPC calls usually keep the same request shape. Update URLs and credentials, then test method limits, tracer options, retries and subscription recovery. Provider-specific methods or node settings may require changes.
How do I choose a production RPC plan?
Measure peak throughput and monthly Response Units from your actual method mix. Check archive and trace access, WebSocket usage, and whether you need isolated node resources.