Defining rolesedit

A role is defined by the following JSON structure:

  "run_as": [ ... ], 
  "cluster": [ ... ], 
  "global": { ... }, 
  "indices": [ ... ], 
  "applications": [ ... ] 


A list of usernames the owners of this role can impersonate.

A list of cluster privileges. These privileges define the cluster level actions users with this role are able to execute. This field is optional (missing cluster privileges effectively mean no cluster level permissions).

An object defining global privileges. A global privilege is a form of cluster privilege that is request sensitive. A standard cluster privilege makes authorization decisions based solely on the action being executed. A global privilege also considers the parameters included in the request. Support for global privileges is currently limited to the management of application privileges. This field is optional.

A list of indices permissions entries. This field is optional (missing indices privileges effectively mean no index level permissions).

A list of application privilege entries. This field is optional.

Role names must be at least 1 and no more than 507 characters. They can contain alphanumeric characters (a-z, A-Z, 0-9), spaces, punctuation, and printable symbols in the Basic Latin (ASCII) block. Leading or trailing whitespace is not allowed.

Indices Privilegesedit

The following describes the structure of an indices permissions entry:

  "names": [ ... ], 
  "privileges": [ ... ], 
  "field_security" : { ... }, 
  "query": "..." 
  "allow_restricted_indices": false 

A list of data streams, indices, and aliases to which the permissions in this entry apply. Supports wildcards (*).

The index level privileges the owners of the role have on the associated data streams and indices specified in the names argument.

Specification for document fields the owners of the role have read access to. See Setting up field and document level security for details.

A search query that defines the documents the owners of the role have read access to. A document within the associated data streams and indices must match this query in order for it to be accessible by the owners of the role.

Restricted indices are a special category of indices that are used internally to store configuration data and should not be directly accessed. Only internal system roles should normally grant privileges over the restricted indices. Toggling this flag is very strongly discouraged because it could effectively grant unrestricted operations on critical data, making the entire system unstable or leaking sensitive information. If however, for administrative purposes, you need to create a role with privileges covering restricted indices, you must set this field to true (default is false), and then the names field will cover the restricted indices as well.

The names parameter accepts wildcard and regular expressions that may refer to multiple data streams, indices, and aliases.

  • Wildcard (default) - simple wildcard matching where * is a placeholder for zero or more characters, ? is a placeholder for a single character and \ may be used as an escape character.
  • Regular Expressions - A more powerful syntax for matching more complex patterns. This regular expression is based on Lucene’s regexp automaton syntax. To enable this syntax, it must be wrapped within a pair of forward slashes (/). Any pattern starting with / and not ending with / is considered to be malformed.

Example Regular Expressions.

"foo-bar":               # match the literal `foo-bar`
"foo-*":                 # match anything beginning with "foo-"
"logstash-201?-*":       # ? matches any one character
"/.*-201[0-9]-.*/":      # use a regex to match anything containing 2010-2019
"/foo":                  # syntax error - missing final /

Global Privilegesedit

The following describes the structure of the global privileges entry:

  "application": {
    "manage": {    
      "applications": [ ... ] 
  "profile": {
    "write": { 
      "applications": [ ... ] 

The privilege for the ability to manage application privileges

The list of application names that may be managed. This list supports wildcards (e.g. "myapp-*") and regular expressions (e.g. "/app[0-9]*/")

The privilege for the ability to write the access and data of any user profile

The list of names, wildcards and regular expressions to which the write privilege is restricted to

Application Privilegesedit

The following describes the structure of an application privileges entry:

  "application": "my_app", 
  "privileges": [ ... ],   
  "resources": [ ... ]     

The name of the application.

The list of the names of the application privileges to grant to this role.

The resources to which those privileges apply. These are handled in the same way as index name pattern in indices permissions. These resources do not have any special meaning to the Elasticsearch security features.

For details about the validation rules for these fields, see the add application privileges API.

A role may refer to application privileges that do not exist - that is, they have not yet been defined through the add application privileges API (or they were defined, but have since been deleted). In this case, the privilege has no effect, and will not grant any actions in the has privileges API.


The following snippet shows an example definition of a clicks_admin role:

POST /_security/role/clicks_admin
  "run_as": [ "clicks_watcher_1" ],
  "cluster": [ "monitor" ],
  "indices": [
      "names": [ "events-*" ],
      "privileges": [ "read" ],
      "field_security" : {
        "grant" : [ "category", "@timestamp", "message" ]
      "query": "{\"match\": {\"category\": \"click\"}}"

Based on the above definition, users owning the clicks_admin role can:

  • Impersonate the clicks_watcher_1 user and execute requests on its behalf.
  • Monitor the Elasticsearch cluster
  • Read data from all indices prefixed with events-
  • Within these indices, only read the events of the click category
  • Within these document, only read the category, @timestamp and message fields.

For a complete list of available cluster and indices privileges

There are two available mechanisms to define roles: using the Role Management APIs or in local files on the Elasticsearch nodes. You can also implement custom roles providers. If you need to integrate with another system to retrieve user roles, you can build a custom roles provider plugin. For more information, see Customizing roles and authorization.

Role management UIedit

You can manage users and roles easily in Kibana. To manage roles, log in to Kibana and go to Management / Security / Roles.

Role management APIedit

The Role Management APIs enable you to add, update, remove and retrieve roles dynamically. When you use the APIs to manage roles in the native realm, the roles are stored in an internal Elasticsearch index. For more information and examples, see Roles.

File-based role managementedit

Apart from the Role Management APIs, roles can also be defined in local roles.yml file located in ES_PATH_CONF. This is a YAML file where each role definition is keyed by its name.

If the same role name is used in the roles.yml file and through the Role Management APIs, the role found in the file will be used.

While the Role Management APIs is the preferred mechanism to define roles, using the roles.yml file becomes useful if you want to define fixed roles that no one (beside an administrator having physical access to the Elasticsearch nodes) would be able to change. Please note however, that the roles.yml file is provided as a minimal administrative function and is not intended to cover and be used to define roles for all use cases.

You cannot view, edit, or remove any roles that are defined in roles.yml by using the role management UI or the role management APIs.

The roles.yml file is managed locally by the node and is not globally by the cluster. This means that with a typical multi-node cluster, the exact same changes need to be applied on each and every node in the cluster.

A safer approach would be to apply the change on one of the nodes and have the roles.yml distributed/copied to all other nodes in the cluster (either manually or using a configuration management system such as Puppet or Chef).

The following snippet shows an example of the roles.yml file configuration:

  run_as: [ 'clicks_watcher_1' ]
  cluster: [ 'monitor' ]
    - names: [ 'events-*' ]
      privileges: [ 'read' ]
        grant: ['category', '@timestamp', 'message' ]
      query: '{"match": {"category": "click"}}'

Elasticsearch continuously monitors the roles.yml file and automatically picks up and applies any changes to it.