Skip to main content

Cluster Datastore

The ability to run Kubernetes using a datastore other than etcd sets K3s apart from other Kubernetes distributions. This feature provides flexibility to Kubernetes operators. The available datastore options allow you to select a datastore that best fits your use case. For example:

  • If your team doesn't have expertise in operating etcd, you can choose an enterprise-grade SQL database like MySQL or PostgreSQL
  • If you need to run a simple, short-lived cluster in your CI/CD environment, you can use the embedded SQLite database
  • If you wish to deploy Kubernetes on the edge and require a highly available solution but can't afford the operational overhead of managing a database at the edge, you can use K3s's embedded HA datastore built on top of embedded etcd.

K3s supports the following datastore options:

  • Embedded SQLite
    SQLite cannot be used on clusters with multiple servers.
    SQLite is the default datastore, and will be used if no other datastore configuration is present, and no embedded etcd database files are present on disk.
  • Embedded etcd
    See the High Availability Embedded etcd documentation for more information on using embedded etcd with multiple servers. Embedded etcd will be automatically selected if K3s is configured to initialize a new etcd cluster, join an existing etcd cluster, or if etcd database files are present on disk during startup.
  • External Database
    See the High Availability External DB documentation for more information on using external datastores with multiple servers.
    The following external datastores are supported:
    • etcd (certified against version 3.5.4)
    • MySQL (certified against versions 5.7 and 8.0)
    • MariaDB (certified against version 10.6.8)
    • PostgreSQL (certified against versions 12.16, 13.12, 14.9 and 15.4)
Prepared Statement Support

K3s requires prepared statements support from the DB. This means that connection poolers such as PgBouncer may require additional configuration to work with K3s.

External Datastore Configuration Parameters

If you wish to use an external datastore such as PostgreSQL, MySQL, or etcd you must set the datastore-endpoint parameter so that K3s knows how to connect to it. You may also specify parameters to configure the authentication and encryption of the connection. The below table summarizes these parameters, which can be passed as either CLI flags or environment variables.

CLI FlagEnvironment VariableDescription
--datastore-endpointK3S_DATASTORE_ENDPOINTSpecify a PostgreSQL, MySQL, or etcd connection string. This is a string used to describe the connection to the datastore. The structure of this string is specific to each backend and is detailed below.
--datastore-cafileK3S_DATASTORE_CAFILETLS Certificate Authority (CA) file used to help secure communication with the datastore. If your datastore serves requests over TLS using a certificate signed by a custom certificate authority, you can specify that CA using this parameter so that the K3s client can properly verify the certificate.
--datastore-certfileK3S_DATASTORE_CERTFILETLS certificate file used for client certificate based authentication to your datastore. To use this feature, your datastore must be configured to support client certificate based authentication. If you specify this parameter, you must also specify the datastore-keyfile parameter.
--datastore-keyfileK3S_DATASTORE_KEYFILETLS key file used for client certificate based authentication to your datastore. See the previous datastore-certfile parameter for more details.

As a best practice we recommend setting these parameters as environment variables rather than command line arguments so that your database credentials or other sensitive information aren't exposed as part of the process info.

Datastore Endpoint Format and Functionality

As mentioned, the format of the value passed to the datastore-endpoint parameter is dependent upon the datastore backend. The following details this format and functionality for each supported external datastore.

In its most common form, the datastore-endpoint parameter for PostgreSQL has the following format:

postgres://username:password@hostname:port/database-name

More advanced configuration parameters are available. For more information on these, please see https://godoc.org/github.com/lib/pq.

If you specify a database name and it does not exist, the server will attempt to create it.

If you only supply postgres:// as the endpoint, K3s will attempt to do the following:

  • Connect to localhost using postgres as the username and password
  • Create a database named kubernetes