Connection errors

This topic describes how to resolve connection errors you may encounter when using Neon. The errors covered include:

• The endpoint ID is not specified
• Password authentication failed for user
• Couldn't connect to compute node
• Can't reach database server
• Error undefined: Database error
• Terminating connection due to administrator command
• Unsupported startup parameter
• You have exceeded the limit of concurrently active endpoints

The endpoint ID is not specified

With older clients and some native Postgres clients, you may receive the following error when attempting to connect to Neon:

ERROR: The endpoint ID is not specified. Either upgrade the Postgres client library (libpq) for SNI support or pass the endpoint ID (the first part of the domain name) as a parameter: '&options=endpoint%3D'. See [https://neon.tech/sni](https://neon.tech/sni) for more information.

This error occurs if your client library or application does not support the Server Name Indication (SNI) mechanism in TLS.

Neon uses computet IDs (the first part of a Neon domain name) to route incoming connections. However, the Postgres wire protocol does not transfer domain name information, so Neon relies on the Server Name Indication (SNI) extension of the TLS protocol to do this.

SNI support was added to libpq (the official Postgres client library) in Postgres 14, which was released in September 2021. Clients that use your system's libpq library should work if your Postgres version is >= 14. On Linux and macOS, you can check Postgres version by running pg_config --version. On Windows, check the libpq.dll version in your Postgres installation's bin directory. Right-click on the file, select Properties > Details.

If a library or application upgrade does not help, there are several workarounds, described below, for providing the required domain name information when connecting to Neon.

A. Pass the endpoint ID as an option

Neon supports a connection option named endpoint, which you can use to identify the compute you are connecting to. Specifically, you can add options=endpoint%3D[endpoint_id] as a parameter to your connection string, as shown in the example below. The %3D is a URL-encoded = sign. Replace [endpoint_id] with your compute's ID, which you can find in your Neon connection string. It looks similar to this: ep-cool-darkness-123456.

postgres://[user]:[password]@[neon_hostname]/[dbname]?options=endpoint%3D[endpoint-id]

The endpoint option works if your application or library permits it to be set. Not all of them do, especially in the case of GUI applications.

B. Use libpq key=value syntax in the database field

If your application or client is based on libpq but you cannot upgrade the library, such as when the library is compiled inside of a an application, you can take advantage of the fact that libpq permits adding options to the database name. So, in addition to the database name, you can specify the endpoint option, as shown below. Replace [endpoint_id] with your compute's endpoint ID, which you can find in your Neon connection string. It looks similar to this: ep-cool-darkness-123456.

dbname=neondb options=endpoint=[endpoint_id]

C. Set verify-full for golang-based clients

If your application or service uses golang Postgres clients like pgx and lib/pg, you can set sslmode=verify-full, which causes SNI information to be sent when you connect. Most likely, this behavior is not intended but happens inadvertently due to the golang's TLS library API design.

D. Specify the endpoint ID in the password field

Another supported workaround involves specifying the endpoint ID in the password field. So, instead of specifying only your password, you provide a string consisting of the endpoint option and your password, separated by a semicolon (;) or dollar sign character ($), as shown in the examples below. Replace [endpoint_id] with your compute's endpoint ID, which you can find in your Neon connection string. It looks similar to this: ep-cool-darkness-123456.

endpoint=<endpoint_id>;<password>

or

endpoint=<endpoint_id>$<password>

Example:

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

This approach causes the authentication method to be downgraded from scram-sha-256 (never transfers a plain text password) to password (transfers a plain text password). However, the connection is still TLS-encrypted, so the level of security is equivalent to the security provided by https websites. We intend deprecate this option when most libraries and applications provide SNI support.

Libraries

Clients on the list of drivers on the PostgreSQL community wiki that use your system's libpq library should work if your libpq version is >= 14.

Neon has tested the following drivers for SNI support:

Password authentication failed for user

The following error is often the result of an incorrectly defined connection information, or the driver you are using does not support Server Name Indication (SNI).

ERROR:  password authentication failed for user '<user_name>' connection to server at "ep-billowing-fun-123456.us-west-2.aws.neon.tech" (12.345.67.89), port 5432 failed: ERROR:  connection is insecure (try using `sslmode=require`)

Check your connection to see if it is defined correctly. Your Neon connection string can be obtained from the Connection Details widget on the Neon Dashboard. It appears similar to this:

postgres://[user]:[password]@[neon_hostname]/[dbname]

For clients or applications that require specifying connection parameters such as user, password, and hostname separately, the values in a Neon connection string correspond to the following:

• User: daniel
• Password: f74wh99w398H
• Hostname: ep-white-morning-123456.us-east-2.aws.neon.tech
• Port number: 5432 (Neon uses default Postgres port, 5432, and is therefore not included in the connection string)
• Database name: neondb (neondb is the ready-to-use database created with each Neon project. Your database name may differ.)

If you find that your connection string is defined correctly, see the instructions regarding SNI support outlined in the preceding section: The endpoint ID is not specified.

Couldn't connect to compute node

This error arises when the Neon proxy, which accepts and handles connections from clients that use the Postgres protocol, fails to establish a connection with your compute. This issue sometimes occurs due to repeated connection attempts during the compute's restart phase after it has been idle due to Autosuspend (scale to zero). Currently, the transition from an idle state to an active one takes a few seconds.

Consider these recommended steps:

• Visit the Neon status page to ensure there are no ongoing issues.
• Pause for a short period to allow your compute to restart, then try reconnecting.
• Try connecting with psql to see if a connection can be established.
• Review the strategies in Connection latency and timeouts for avoiding connection issues due to compute startup time.

If the connection issue persists, please reach out to Support.

Can't reach database server

This error is sometimes encountered when using Prisma Client with Neon.

Error: P1001: Can't reach database server at `ep-white-thunder-826300.us-east-2.aws.neon.tech`:`5432`
Please make sure your database server is running at `ep-white-thunder-826300.us-east-2.aws.neon.tech`:`5432`.

A compute in Neon has two main states: Active and Idle. Active means that Postgres is currently running. If there are no active queries for 5 minutes, the activity monitor gracefully places the compute into an idle state to reduce compute usage.

When you connect to an idle compute, Neon automatically activates it. Activation typically happens within a few seconds. If the error above is reported, it most likely means that the Prisma query engine timed out before your Neon compute was activated. For dealing with this connection timeout scenario, refer to the connection timeout instructions in our Prisma documentation. Our connection latency and timeout documentation may also be useful in addressing this issue.

Error undefined: Database error

This error is sometimes encountered when using Prisma Migrate with Neon.

Error undefined: Database error
Error querying the database: db error: ERROR: prepared statement
"s0" already exists

Prisma Migrate requires a direct connection to the database. It does not support a pooled connection with PgBouncer, which is the connection pooler used by Neon. Attempting to run Prisma Migrate commands, such as prisma migrate dev, with a pooled connection causes this error. To resolve this issue, please refer to our Connection pooling with Prisma Migrate instructions.

Terminating connection due to administrator command

The terminating connection due to administrator command error is typically encountered when running a query from a connection that has sat idle long enough for the compute to suspend due to inactivity. Neon automatically suspends a compute after 5 minutes of inactivity, by default. You can reproduce this error by connecting to your database from an application or client such as psql, letting the connection remain idle until the compute suspends, and then running a query from the same connection.

If you encounter this error, you can try adjusting the timing of your query or reestablishing the connection before running the query. Alternatively, if you are a paying user, you can disable autosuspend or configure a different suspension period. For instructions, see Configuring Autosuspend for Neon computes. Neon Free Tier users cannot modify the default 5 minute autosuspend setting.

Unsupported startup parameter

This error is reported in two variations:

unsupported startup parameter: <...>

unsupported startup parameter in options: <...>

The error occurs when using a pooled Neon connection string with startup options that are not supported by PgBouncer. PgBouncer allows only startup parameters it can keep track of in startup packets. These include: client_encoding, datestyle, timezone, standard_conforming_strings, and application_name. See track_extra_parameters, in the PgBouncer documentation. To resolve this error, you can either remove the unsupported parameter from your connection string or use an unpooled Neon connection string. For information about pooled and unpooled connections in Neon, see Connection pooling.

You have exceeded the limit of concurrently active endpoints

This error can also appear as: active endpoints limit exceeded.

Neon has a default limit of 20 concurrently active computes to protect your account from unintended usage. The compute associated with the default branch is exempt from this limit, ensuring that it is always available. When you exceed the limit, any compute associated with a non-default branch will remain suspended and you will see this error when attempting to connect to it. You can suspend computes and try again. Alternatively, if you encounter this error often, you can reach out to Support to request a limit increase.