Getting ready for production

Set minimum compute to at least 1 CU

Before your application goes to prodution, make sure your database has enough vCPU and memory to handle expected production load. See How to size your compute.

Recommendation

We recommend that you fit your data in memory and use Neon autoscaling:

• Start with a compute size that can hold all your data in memory. Or try to fit at least your most frequently accessed data (your working set).
• Once you determine the right size for your compute, use that as the minimum compute size for Autoscaling).

About compute size

A Compute Unit (CU) in Neon measures the processing power or "size" of a Neon compute. One CU includes 1 vCPU and 4 GB of RAM. Neon computes can range from 0.25 CUs to 56 CUs, depending on your Neon plan.

Set max compute to the highest CU available for your plan

Use Neon's autoscaling feature to dynamically adjust your compute resources based on your current workload. This means you don't need to scale manually during traffic surges.

Recommendation

• Minimum compute size: Autoscaling works best if your data can be fully cached in memory.
• Maximum compute size: Set this to as a high a limit as your plan allows. You only pay for what you use.

To get started with Autoscaling, read:

• Enable Autoscaling in Neon
• How to size your compute, including the Autoscaling considerations section.

Disable scale to zero

Scale to zero turns off your compute after a period of inactivity. Ideal for development or other environments with bursty usage.

Recommendation

Disable scale to zero for production workloads. This ensures your compute is always active, preventing delays and session context resets caused by cold starts.

Session and latency considerations

By default, your compute scales to zero after 5 minutes. Restarts are nearly instant, but there may still be some latency (around 500 milliseconds depending on the region). Restarts will reset your session context, affecting in-memory statistics and temporary tables. While typical production loads might never go idle long enough to scale to zero, disabling this feature prevents any possible issues from cold starts or session loss.

Disabling scale to zero is available on paid plans only. See Configuring Scale to Zero for Neon computes for more detail.

Use a pooled connection

Connection pooling with PgBouncer allows your database to handle up to 10,000 concurrent connections, reducing connection overhead and improving performance.

Recommendation

For production environments, enable connection pooling. This increases the number of simultaneous connections your database can handle and optimizes resource usage.

Connection details

Use a pooled connection string by adding -pooler to your endpoint ID, or simply copy the pooled connection string from the Connect widget in your Project Dashboard. Use this string as your database connection in your application's environment variables. For more information, see Connection pooling.

Example connection string:

postgresql://alex:AbC123dEf@ep-cool-darkness-123456-pooler.us-east-2.aws.neon.tech/dbname?sslmode=require

Increase your project's restore window to 7 days

Neon retains a history of changes for all branches, enabling instant restore and time travel queries. This history acts as a backup strategy, allowing recovery of lost data and viewing past database states.

Recommendation

Set your restore window to 7 days to ensure data integrity and quick recovery.

Restore window details

By default, Neon's restore window is set to 1 day. Extending it to 7 days helps protect you against data loss, letting you recover from human or application errors that may go unnoticed for days. It can also help you comply with any industry regulations that need longer retention periods. While a longer restore window can increase storage costs, it provides exta security and recoverability for production data.

For more info, see Instant restore.

Restrict database access to trusted IPs

Neon's IP Allow feature ensures that only trusted IP addresses can connect to your database, preventing unauthorized access and enhancing security.

Recommendation

Combine an allowlist with protected branches for enhanced security. This setup ensures that only trusted IPs can access critical data, reducing the risk of unauthorized access and safeguarding data integrity.

Configuration details

• IP Allow: Restricts access to specific, trusted IP addresses, preventing unauthorized connections.
• Protected branch: Safeguards critical data from accidental deletions or modifications by designating branches as protected.

Available with the Neon Scale and Business plans, you can configure IP Allow and protected branches in your Neon project's settings. For more information, see Configure IP Allow and Protected branches guide.

Set up metrics export to Datadog

Export Neon metrics to DataDog and centralize your database monitoring with your existing observability stack.

Recommendation

Set up Datadog integration to monitor and set alerts for key metrics:

• Connection counts (active and idle database connections)
• Database size (total size of all databases)
• Replication delay (lag in bytes and seconds)
• Compute metrics (CPU and memory usage)

For more information, see The Neon Datadog integration.

Ensure your app reconnects after your database restarts

Verify your application handles compute restarts gracefully. Neon occasionally restarts computes for updates and maintenance.

Recommendation

Most database drivers and connection pools handle reconnection automatically, but it's important to test this behavior. You can use the Neon API to trigger a restart and watch your application reconnect:

curl --request POST \
  --url https://console.neon.tech/api/v2/projects/your_project_id/endpoints/your_endpoint_id/restart \
  --header 'accept: application/json' \
  --header 'authorization: Bearer $NEON_API_KEY'

See Restart compute endpoint for details.

For more information:

• Build connection timeout handling into your application
• Maintenance & updates overview

Install pg_stat_statements

Enable query performance monitoring to track execution times and frequency.

Recommendation

Install the pg_stat_statements extension to monitor query performance and identify potential bottlenecks.

Usage

CREATE EXTENSION IF NOT EXISTS pg_stat_statements;

The statistics gathered by this extension require little overhead and let you quickly access metrics like:

• Most frequently executed queries
• Longest running queries
• Queries that return the most rows

You can also use the Monitoring Dashboard in the Neon Console to view live graphs for system and database metrics like CPU, RAM, and connections.

For more information, see Query performance and Monitoring.

Upgrade to a Neon Business plan for priority support

Support tickets opened by Business and Enterprise support plan customers are given top priority by the Neon Support team.

Recommendation

Upgrade to a Business plan to get both Priority support and acccess to the Business SLA.

For more information, see the Support documentation.