API keys

Configure API keys to authorize requests to the APM Server. To enable API key authorization, set apm-server.auth.api_key.enabled to true.

There are multiple, unique privileges you can assign to each API key. API keys can have one or more of these privileges:

To secure the communication between APM Agents and the APM Server with API keys, make sure SSL/TLS is enabled, then complete these steps:

API keys are disabled by default. Enable and configure this feature in the apm-server.auth.api_key section of the apm-server.yml configuration file.

At a minimum, you must enable API keys, and should set a limit on the number of unique API keys that APM Server allows per minute. Here’s an example apm-server.auth.api_key config using 50 unique API keys:

apm-server.auth.api_key.enabled: true 
apm-server.auth.api_key.limit: 50 

All other configuration options are described in API key configuration options.

API keys can only have the same or lower access rights than the user that creates them. Instead of using a superuser account to create API keys, you can create a role with the minimum required privileges.

The user creating an APM agent API key must have at least the manage_own_api_key cluster privilege and the APM application-level privileges that it wishes to grant. The example below uses the Kibana role management API to create a role named apm_agent_key_role.

POST /_security/role/apm_agent_key_role
{
  "cluster": ["manage_own_api_key"],
  "applications": [{
    "application": "apm",
    "privileges": ["event:write", "config_agent:read", "sourcemap:write"],
    "resources": ["*"]
  }]
}

Assign the newly created apm_agent_key_role role to any user that wishes to create APM agent API keys.

Using a superuser account, or a user with the role created in the previous step, open Kibana, navigate to Stack Management > API keys and click Create API key.

Enter a name for your API key and select Restrict privileges. In the role descriptors box, copy and paste the following JSON. This example creates an API key with privileges for ingesting APM events, reading agent central configuration, uploading a source map:

{
   "apm": {
      "applications": [
         {
            "application": "apm",
            "privileges": ["sourcemap:write", "event:write", "config_agent:read"], 
            "resources": ["*"]
         }
      ]
   }
}

To set an expiration date for the API key, select Expire after time and input the lifetime of the API key in days.

Click Create API key and then copy the Base64 encoded API key. You will need this for the next step, and you will not be able to view it again.

You can now apply your newly created API keys in the configuration of each of your APM agents. See the relevant agent documentation for additional information:

API keys can also be created and validated outside of Kibana:

APM Server provides a command line interface for creating, retrieving, invalidating, and verifying API keys. Keys created using this method can only be used for communication with APM Server.

If privileges are not specified at creation time, the created key will have all privileges.

Create an API key with the create subcommand.

The following example creates an API key with a name of java-001, and gives the "agent configuration" and "ingest" privileges.

apm-server apikey create --ingest --agent-config --name java-001

The response will look similar to this:

Name ........... java-001
Expiration ..... never
Id ............. qT4tz28B1g59zC3uAXfW
API Key ........ rH55zKd5QT6wvs3UbbkxOA (won't be shown again)
Credentials .... cVQ0dHoyOEIxZzU5ekMzdUFYZlc6ckg1NXpLZDVRVDZ3dnMzVWJia3hPQQ== (won't be shown again)

You should always verify the privileges of an API key after creating it. Verification can be done using the verify subcommand.

The following example verifies that the java-001 API key has the "agent configuration" and "ingest" privileges.

apm-server apikey verify --agent-config --ingest --credentials cVQ0dHoyOEIxZzU5ekMzdUFYZlc6ckg1NXpLZDVRVDZ3dnMzVWJia3hPQQ==

If the API key has the requested privileges, the response will look similar to this:

Authorized for privilege "event:write"...:          Yes
Authorized for privilege "config_agent:read"...:    Yes

To invalidate an API key, use the invalidate subcommand. Due to Elasticsearch caching, there may be a delay between when this subcommand is executed and when it takes effect.

The following example invalidates the java-001 API key.

apm-server apikey invalidate --name java-001

The response will look similar to this:

Invalidated keys ... qT4tz28B1g59zC3uAXfW
Error count ........ 0

A full list of apikey subcommands and flags is available in the API key command reference.

It is also possible to create API keys using the Elasticsearch create API key API.

This example creates an API key named java-002:

POST /_security/api_key
{
  "name": "java-002", 
  "expiration": "1d", 
  "role_descriptors": {
    "apm": {
      "applications": [
        {
          "application": "apm",
          "privileges": ["sourcemap:write", "event:write", "config_agent:read"], 
          "resources": ["*"]
        }
      ]
    }
  }
}

The response will look similar to this:

{
  "id" : "GnrUT3QB7yZbSNxKET6d",
  "name" : "java-002",
  "expiration" : 1599153532262,
  "api_key" : "RhHKisTmQ1aPCHC_TPwOvw"
}

The credential string, which is what agents use to communicate with APM Server, is a base64 encoded representation of the API key’s id:api_key. It can be created like this:

echo -n GnrUT3QB7yZbSNxKET6d:RhHKisTmQ1aPCHC_TPwOvw | base64

You can verify your API key has been base64-encoded correctly with the Authenticate API:

curl -H "Authorization: ApiKey R0gzRWIzUUI3eVpiU054S3pYSy06bXQyQWl4TlZUeEcyUjd4cUZDS0NlUQ==" localhost:9200/_security/_authenticate

If the API key has been encoded correctly, you’ll see a response similar to the following:

{
   "username":"1325298603",
   "roles":[],
   "full_name":null,
   "email":null,
   "metadata":{
      "saml_nameid_format":"urn:oasis:names:tc:SAML:2.0:nameid-format:transient",
      "saml(http://saml.elastic-cloud.com/attributes/principal)":[
         "1325298603"
      ],
      "saml_roles":[
         "superuser"
      ],
      "saml_principal":[
         "1325298603"
      ],
      "saml_nameid":"_7b0ab93bbdbc21d825edf7dca9879bd8d44c0be2",
      "saml(http://saml.elastic-cloud.com/attributes/roles)":[
         "superuser"
      ]
   },
   "enabled":true,
   "authentication_realm":{
      "name":"_es_api_key",
      "type":"_es_api_key"
   },
   "lookup_realm":{
      "name":"_es_api_key",
      "type":"_es_api_key"
   }
}

You can then use the APM Server CLI to verify that the API key has the requested privileges:

apm-server apikey verify --credentials R25yVVQzUUI3eVpiU054S0VUNmQ6UmhIS2lzVG1RMWFQQ0hDX1RQd092dw==

If the API key has the requested privileges, the response will look similar to this:

Authorized for privilege "config_agent:read"...:  Yes
Authorized for privilege "event:write"...:        Yes
Authorized for privilege "sourcemap:write"...:    Yes

You can specify the following options in the apm-server.auth.api_key.* section of the apm-server.yml configuration file. They apply to API key communication between the APM Server and APM Agents.

Enable API key authorization by setting enabled to true. By default, enabled is set to false, and API key support is disabled.

Each unique API key triggers one request to Elasticsearch. This setting restricts the number of unique API keys are allowed per minute. The minimum value for this setting should be the number of API keys configured in your monitored services. The default limit is 100.

All of the auth.api_key.elasticsearch.* configurations are optional. If none are set, configuration settings from the apm-server.output section will be reused.

API keys are fetched from Elasticsearch. This configuration needs to point to a secured Elasticsearch cluster that is able to serve API key requests.

The name of the protocol Elasticsearch is reachable on. The options are: http or https. The default is http. If nothing is configured, configuration settings from the output section will be reused.

An optional HTTP path prefix that is prepended to the HTTP API calls. If nothing is configured, configuration settings from the output section will be reused.

The URL of the proxy to use when connecting to the Elasticsearch servers. The value may be either a complete URL or a "host[:port]", in which case the "http"scheme is assumed. If nothing is configured, configuration settings from the output section will be reused.

The HTTP request timeout in seconds for the Elasticsearch request. If nothing is configured, configuration settings from the output section will be reused.

SSL is off by default. Set elasticsearch.protocol to https if you want to enable https.

Enable custom SSL settings. Set to false to ignore custom SSL settings for secure communication.

Configure SSL verification mode. If none is configured, all server hosts and certificates will be accepted. In this mode, SSL based connections are susceptible to man-in-the-middle attacks. Use only for testing. Default is full.

List of supported/valid TLS versions. By default, all TLS versions from 1.0 to 1.2 are enabled.

List of root certificates for HTTPS server verifications.

The path to the certificate for SSL client authentication.

The client certificate key used for client authentication. This option is required if certificate is specified.

An optional passphrase used to decrypt an encrypted key stored in the configured key file. It is recommended to use the provided keystore instead of entering the passphrase in plain text.

The list of cipher suites to use. The first entry has the highest priority. If this option is omitted, the Go crypto library’s default suites are used (recommended).

The list of curve types for ECDHE (Elliptic Curve Diffie-Hellman ephemeral key exchange).

Configure what types of renegotiation are supported. Valid options are never, once, and freely. Default is never.

In versions prior to 7.14.0, API Key authorization was known as apm-server.api_key. In 7.14.0 this was renamed apm-server.auth.api_key. The old configuration will continue to work until 8.0.0, and the new configuration will take precedence.