# Connections and timeouts

Configure client lifecycles, pool sizes and timeouts for a ParticleDB application.

Source: https://particledb.ai/operations/connections/

## Verify the address

Use the configured [client address](/installation/ports/), database and credentials. A web console URL is not a database connection string.

## Use a bounded pool

Reuse connections where the client supports pooling. Choose a pool size that matches application concurrency and server capacity. Close connections when the application stops.

Keep a transaction on one checked-out connection until commit or rollback. A connection returned to the pool with an open or aborted transaction can affect the next request.

## Set timeouts

Distinguish connection establishment, network reads, SQL execution and idle sessions. Configure each through the client or server interface that supports it. A timeout does not prove a write was not committed; reconcile uncertain results using stable identifiers.

## Recover

After a broken connection, create a new connection. After a transaction conflict, retry the transaction using a bounded policy. After authentication failure, fix credentials rather than retrying indefinitely.

[Client guides](/sdks/overview/) · [Troubleshooting](/guides/troubleshooting/)

## Related

[Documentation home](/docs/) · [Troubleshooting](/guides/troubleshooting/)