Read Replicas

Read replicas are supported with all SerenDB plans. The Free plan is limited to a maximum of 3 read replica computes per project. This guide steps you through the process of creating and managing read replicas.

The general purpose of read replicas is to segregate read-only work from your production database operations. This can be applied to different uses cases, such as:

Horizontal scaling: Distributing read requests across replicas to improve performance and increase throughput
Analytics queries: Offloading resource-intensive analytics and reporting workloads to reduce load on the primary compute
Read-only access: Granting read-only access to users or applications that don't require write permissions

Regardless of the application, the steps for creating, configuring, and connecting to a read replica are the same. You can create one or more read replicas for any branch in your SerenDB project and configure the vCPU and memory allocated to each. SerenDB's Autoscaling and Scale to Zero features are also supported, providing you with control over read replica compute usage.

Prerequisites

A SerenDB account
A SerenDB project

Create a read replica

Creating a read replica involves adding a read replica compute to a branch. You can add a read replica compute to any branch in your SerenDB project using the SerenDB Console, SerenDB CLI, or SerenDB API.

The Free plan is limited to a maximum of 3 read replica computes per project.

To create a read replica from the SerenDB Console:

In the SerenDB Console, select Branches.
Select the branch where your database resides.
Click Add Read Replica.
On the Add new compute dialog, select Read replica as the Compute type.
Specify the Compute size settings. You can configure a Fixed Size compute with a specific amount of vCPU and RAM (the default) or enable autoscaling by configuring a minimum and maximum compute size. You can also configure the Suspend compute after inactivity setting, which is the amount of idle time after which your compute is automatically suspended. The default setting is 5 minutes. The compute size configuration determines the processing power of your database.
When you finish making your selections, click Create.

In a few seconds, your read replica is provisioned and appears on the Computes tab on the Branches page. The following section describes how to connect to your read replica.

To create a read replica using the SerenDB CLI, use the branches command, specifying the add-compute subcommand with --type read_only. If you have more than one SerenDB project, also include the --project-id option.

neon branches add-compute mybranch --type read_only

To create a read replica compute using the SerenDB API, use the Create endpoint method. The type attribute in the following example specifies read_only, which creates a read replica compute. For information about obtaining the required project_id and branch_id parameters, refer to Create an endpoint, in the Neon API reference.

curl --request POST \
     --url https://console.serendb.com/api/v2/projects/<project_id>/endpoints \
     --header 'Accept: application/json' \
     --header "Authorization: Bearer $NEON_API_KEY" \
     --header 'Content-Type: application/json' \
     --data '
{
  "endpoint": {
    "type": "read_only",
    "branch_id": "<branch_id>"
  }
}
' | jq

Connect to a read replica

Connecting to a read replica is the same as connecting to any branch, except you connect via a read replica compute instead of your primary read-write compute. The following steps describe how to connect to your read replica with connection details obtained from the SerenDB Console.

Click the Connect button on your Project Dashboard. On the Connect to your database modal, select the branch, the database, and the role you want to connect with.
Under Compute, select a Replica.
Select a connection string or a code example from the drop-down menu and copy it. This is the information you need to connect to the read replica from your client or application.
A psql connection string appears similar to the following:
```
postgresql://[user]:[password]@[seren_hostname]/[dbname]?sslmode=require&channel_binding=require
```
If you expect a high number of connections, enable the Connection pooling toggle to add the -pooler flag to the connection string or example.
Write operations are not permitted on a read replica connection.

View read replicas

You can view read replicas using the SerenDB Console or SerenDB API.

To view read replicas for a branch, select **Branches** in the SerenDB Console, and select a branch. Read replicas are listed on the **Computes** tab.

To view read replica computes with the [SerenDB API](https://api-docs.serendb.com/reference/createprojectendpoint), use the [Get endpoints](https://api-docs.serendb.com/reference/listprojectendpoints) method.

curl -X 'GET' \
  'https://console.serendb.com/api/v2/projects/<project_id>/endpoints' \
  -H 'accept: application/json' \
  -H "Authorization: Bearer $NEON_API_KEY"

For information about obtaining the required project_id parameter for this command, refer to Get endpoints, in the Neon API reference. For information about obtaining an SerenDB API key, see Create an API key.

In the response body for this method, read replica computes are identified by the type value, which is read_only.

Edit a read replica

You can edit a read replica using the SerenDB Console or SerenDB API to change the Compute size or Scale to Zero configuration.

To edit a read replica compute using the SerenDB Console:

In the SerenDB Console, select Branches.
Select a branch.
Under Computes, identify the read replica compute you want to modify, and click Edit.
Make the changes to your compute settings, and click Save.

To edit a read replica compute with the SerenDB API, use the [Update endpoint](https://api-docs.serendb.com/reference/updateprojectendpoint) method.

curl --request PATCH \
     --url https://console.serendb.com/api/v2/projects/<project_id>/endpoints/<endpoint_id> \
     --header 'Accept: application/json' \
     --header "Authorization: Bearer $NEON_API_KEY" \
     --header 'Content-Type: application/json' \
     --data '
{
  "endpoint": {
    "autoscaling_limit_min_cu": 25,
    "autoscaling_limit_max_cu": 3,
    "suspend_timeout_seconds": 604800,
    "provisioner": "k8s-neonvm"
  }
}
'

Computes are identified by their project_id and endpoint_id. For information about obtaining the required project_id and endpoint_id parameters, refer to Update endpoint, in the Neon API reference. For information about obtaining an SerenDB API key, see Create an API key.

Delete a read replica

You can delete a read replica using the SerenDB Console or SerenDB API. Deleting a read replica is a permanent action, but you can quickly create a new read replica if you need one.

To delete a read replica using the SerenDB Console:

In the SerenDB Console, select Branches.
Select a branch.
On the Computes tab, find the read replica you want to delete.
Click Edit → Delete.

To delete a read replica compute with the SerenDB API, use the [Delete endpoint](https://api-docs.serendb.com/reference/deleteprojectendpoint) method.

curl --request DELETE \
     --url https://console.serendb.com/api/v2/projects/<project_id>/endpoints/<endpoint_id> \
     --header 'Accept: application/json' \
     --header "Authorization: Bearer $NEON_API_KEY"

Computes are identified by their project_id and endpoint_id. For information about obtaining the required project_id and endpoint_id parameters, refer to Delete endpoint, in the Neon API reference. For information about obtaining an SerenDB API key, see Create an API key.

Monitoring read replicas

You can monitor replication delay between the primary compute and your read replica computes from the Monitoring page in the SerenDB Console. Two graphs are provided:

Replication delay bytes

The Replication delay bytes graph shows the total size, in bytes, of the data that has been sent from the primary compute but has not yet been applied on the replica. A larger value indicates a higher backlog of data waiting to be replicated, which may suggest issues with replication throughput or resource availability on the replica. This graph is only visible when selecting a Replica compute from the Compute drop-down menu.

Replication delay seconds

The Replication delay seconds graph shows the time delay, in seconds, between the last transaction committed on the primary compute and the application of that transaction on the replica. A higher value suggests that the replica is behind the primary, potentially due to network latency, high replication load, or resource constraints on the replica. This graph is only visible when selecting a Replica compute from the Compute drop-down menu.

Read replica compute setting synchronization

For SerenDB read replicas, certain Postgres settings should not have lower values than your primary read-write compute. For this reason, the following settings on read replica computes are synchronized with the settings on the primary read-write compute when the read replica compute is started:

max_connections
max_prepared_transactions
max_locks_per_transaction
max_wal_senders
max_worker_processes

No users action is required. The settings are synchronized automatically when you create a read replica. However, if you change the compute size configuration on the primary read-write compute, you will need to restart your read replica computes to ensure that settings remain synchronized, as described in the next section.

Replication delay issues

If your read replicas are falling behind, follow these steps to diagnose and resolve the issue:

Check your replication lag metrics Refer to Monitoring Read Replicas for instructions on how to monitor replication lag.
Verify configuration alignment If replication lag is detected, ensure that the configurations for the primary and read-replica computes are aligned. Specifically, confirm that the following parameters match between your primary compute and read-replica compute:
- max_connections
- max_prepared_transactions
- max_locks_per_transaction
- max_wal_senders
- max_worker_processes
Restart read-replica computes if configurations are misaligned If the configurations are not aligned, restart your read-replica computes to automatically update their settings. For instructions, see Restart a Compute.
When increasing the size of your primary read-write compute, always restart associated read replicas to ensure their configurations remain aligned.

PreviousAutoscaling Guide NextBackup & Restore

Last updated 1 month ago