Designer configuration
Web interface configuration
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
http.port | High | int | 8080 | HTTP Port of Designer app |
http.interface | High | string | "0.0.0.0" | HTTP interface for Designer app |
http.publicPath | Medium | string | "" | if Designer used with reverse proxy and custom path, use this configuration to generate links in Designer properly (Designer app is always served from root path) |
ssl.enabled | Medium | boolean | false | Should Designer app be served with SSL |
ssl.keyStore.location | Medium | string | Keystore file location (required if SSL enabled) | |
ssl.keyStore.password | Medium | string | Keystore file password (required if SSL enabled) | |
akka.* | Medium | Akka HTTP is used for HTTP serving, you can configure it in standard way. Below we give Nussknacker-specific defaults | ||
akka.http.server.parsing.max-content-length | Low | int | 300000000 | Requests (e.g. with test data) can be quite large, so we increase the limit |
akka.http.server.request-timeout | Low | duration | 1 minute | Consider increasing the value if you have large test data or long savepoint times during deploy |
Database configuration
Currently, Nussknacker supports following databases:
- HSQL (embedded), we use
syntax_ora
option - PostgreSQL
Please check Slick documentation for detailed list of configuration options.
The table below presents most important options, or the ones that have Nussknacker specific defaults.
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
db.url | High | string | "jdbc:hsqldb:file:"${storageDir}"/db;sql.syntax_ora=true" | Default HSQL location |
db.driver | High | string | "org.hsqldb.jdbc.JDBCDriver" | |
db.user | High | string | "SA" | |
db.password | High | string | "" | |
db.connectionTimeout | Low | int | 30000 | |
db.maximumPoolSize | Low | int | 5 | We have lower limits than default config, since then Designer is not heavy-load application |
db.minimumIdle | Low | int | 1 | We have lower limits than default config, since then Designer is not heavy-load application |
db.numThreads | Low | int | 5 | We have lower limits than default config, since then Designer is not heavy-load application |
Metrics settings
Metric dashboard
Each scenario can have a link to grafana dashboard. In docker setup we
provide nussknacker-scenario
dashboard.
You can modify/configure own, the only assumption that we make is that variable scenarioName
is used to display metrics for particular scenario.
Each scenario type can have different dashboard, this is configured by
metricsSettings.scenarioTypeToDashboard
settings. If no mapping is configured, metricsSettings.defaultDashboard
is used.
Actual link for particular scenario is created by replacing
$dashboard
with configured dashboard$scenarioName
with scenario name inmetricsSettings.url
setting.
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
metricsSettings.url | High | string | /grafana/d/$dashboard?theme=dark&var-scenarioName=$scenarioName&var-env=local (for docker setup) | URL (accessible from user browser, in docker setup its configured as relative URL) to Grafana dashboard, see above for details |
metricsSettings.defaultDashboard | Medium | string | nussknacker-scenario (for docker setup) | Default dashboard |
metricsSettings.scenarioTypeToDashboard | Low | map | Mapping of scenario types to dashboard |
Counts
Counts are based on InfluxDB metrics, stored in nodeCount
measurement by default.
countsSettings.queryMode
setting can be used to choose metric computation algorithm:
OnlySingleDifference
- subtracts values between end and beginning of requested range. Fast method, but if restart in the requested time range is detected error is returned. We assume the job was restarted when event counter at the source decreases.OnlySumOfDifferences
- difference is computed by summing differences in measurements for requested time range. This method works a bit better for restart situations, but can be slow for large diagrams and wide time ranges.SumOfDifferencesForRestarts
- if restart is detected, the metrics are computed withOnlySumDifferences
, otherwise - withOnlySingleDifferences
If you have custom metrics settings which result in different fields or tags (e.g. you have different telegraf configuration), you can configure required values with the settings presented below:
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
countsSettings.influxUrl | Medium | string | Main InfluxDB query endpoint (e.g. http://influx:8086/query). It should be accessible from Nussknacker Designer server, not from user browser | |
countsSettings.database | Medium | string | ||
countsSettings.user | Medium | string | ||
countsSettings.password | Medium | string | ||
countsSettings.queryMode | Low | OnlySingleDifference / OnlySumOfDifferences / SumOfDifferencesForRestarts | SumOfDifferencesForRestarts | |
countsSettings.metricsConfig.sourceCountMetric | Low | string | source_count | |
countsSettings.metricsConfig.nodeCountMetric | Low | string | nodeCount | |
countsSettings.metricsConfig.nodeIdTag | Low | string | nodeId | |
countsSettings.metricsConfig.additionalGroupByTags | Low | list of string | ["slot", "instanceId"] | |
countsSettings.metricsConfig.scenarioTag | Low | string | scenario | |
countsSettings.metricsConfig.countField | Low | string | count | |
countsSettings.metricsConfig.envTag | Low | string | env |
Deployment settings
Nussknacker Designer can be configured to replace certain values in comments to links that can point e.g. to external issue tracker like
GitHub issues or Jira. For example, MARKETING-555
will change to link https://jira.organization.com/jira/browse/MARKETING-555
.
See development configuration for example configuration.
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
commentSettings.matchExpression | Low | regexp | Regular expression to look for issue identifier (e.g. (issues/[0-9]*) - note use of regexp group) | |
commentSettings.link | Low | string | Link template (e.g. https://github.com/TouK/nussknacker/$1 - $1 will be replaced with matched group from matchExpression config | |
deploySettings.requireComment | Low | boolean | false | If true, comment is required for deployment. Also, if matchExpression is defined, at least one match is required |
Security
Nussknacker has pluggable security architecture - by default we support two type of authentication: BasicAuth and OAuth2. You can either use default authentication provider, based on Basic authentication and static user configuration or integrate with other authentication mechanisms such as custom SSO implementation.
Users and permissions
Each user has id and set of permissions for every scenario category. There are following permissions:
- Read - user can view scenarios in category
- Write - user can modify/add new scenarios in category
- Deploy - user can deploy or cancel scenarios in given category
You can set a role assigned to an anonymous user with the anonymousUserRole
setting in the authentication
section in the configuration.
When no value is provided (default), no anonymous access will be granted.
Global permissions
In addition to permission system oriented around scenarios categories we provide additional set of permissions. This feature is designed to control access to components that have no category attached, or it doesn't make sense for them to have one.
Currently supported permissions:
- AdminTab - shows Admin tab in the UI (right now there are some useful things kept there including search components functionality).
Configuration parameters
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
authentication.method | required | string | BasicAuth , Oidc or OAuth2 | |
authentication.usersFile | required | url or path | URL or path to a file with a mapping of user identities to roles and roles to permissions | |
authentication.anonymousUserRole | optional | string | Role assigned to an unauthenticated user if the selected authentication provider permits anonymous access. No anonymous access allowed unless a value is provided. |
Users' file format:
users: [
{
identity: "admin"
roles: ["Admin"]
},
{
identity: "userWithAdminTab"
roles: ["User", "UserWithAdminTab"]
}
{
identity: "user"
roles: ["User"]
}
]
rules: [
{
role: "Admin"
isAdmin: true,
categories: ["RequestResponseCategory1"]
},
{
role: "UserWithAdminTab"
permissions: ["Read", "Write", "Deploy"]
globalPermissions: ["AdminTab"]
categories: ["Category2", "RequestResponseCategory1"]
},
{
role: "User"
permissions: ["Read", "Write"]
categories: ["Category1", "Category2"]
}
]
BasicAuth security module
In order to use Basic authentication you just need to set authentication.method
to BasicAuth
and
provide either plain or encrypted passwords for users additionally to the usersFile
's content as follows:
users: [
{
...
password: "pass"
},
{
...
encryptedPassword: "$2a$12$oA3U7DXkT5eFkyB8GbtKzuVqxUCU0zDmcueBYV218zO/JFQ9/bzY6"
}
]
...
Encrypted hashes
Encryption of passwords uses BCrypt algorithm. You can generate sample hash using command:
python -c 'import bcrypt; print(bcrypt.hashpw("password".encode("utf8"), bcrypt.gensalt(rounds=12, prefix="2a")))'
If you don't have bcrypt installed, use pip install bcrypt
.
Be aware that usage of BCrypt hashes will cause significant CPU overhead for processing of each http request, because we don't have sessions and all requests are authenticated. To avoid this overhead you can configure cashing of hashes using configuration:
authentication: {
method: "BasicAuth"
usersFile: "conf/users.conf"
cachingHashes {
enabled: true
}
}
This workaround causes that passwords are kept in the memory, and it will introduce risk that someone with access to content of heap will see cached passwords.
OpenID Connect security module
When talking about OAuth2 in the context of authentication, most people probably mean OpenID Connect, an identity layer built on top of it. Nussknacker provides a separate authentication provider for OIDC with simple configuration and provider discovery. The only supported flow is the authorization code flow with client secret.
You can select this authentication method by setting the authentication.method
parameter to Oidc
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
authentication.issuer | required | url | OpenID Provider's location | |
authentication.clientId | required | string | Client identifier valid at the authorization server | |
authentication.clientSecret | required | string | Secret corresponding to the client identifier at the authorization server | |
authentication.audience | recommended | string | Required aud claim value of an access token that is assumed to be a JWT. | |
authentication.rolesClaims | recommended | list of string | ID Token claims used for mapping users to roles instead of or additionally to the ones defined in usersFile | |
authentication.redirectUri | optional | url | inferred from UI's location | Callback URL to which a user is redirected after successful authentication |
authentication.scope | optional | string | openid profile | Scope parameter's value sent to the authorization endpoint. |
authentication.authorizationEndpoint | auxiliary | url or path | discovered | Absolute URL or path relative to Issuer overriding the value retrieved from the OpenID Provider |
authentication.tokenEndpoint | auxiliary | url or path | discovered | as above |
authentication.userinfoEndpoint | auxiliary | url or path | discovered | as above |
authentication.jwksUri | auxiliary | url or path | discovered | as above |
Auth0 sample configuration
Assuming ${nussknackerUrl}
is the location of your deployment, in your Auth0 tenant settings do the following:
- Create a "Regular Web Application" with the "Allowed Callback URL's" field set to
${nussknackerUrl}
- Create an "API" with the "Identifier" field preferably set to
${nussknackerUrl}/api
- Create an Auth Pipeline Rule with the content:
function (user, context, callback) {
const assignedRoles = (context.authorization || {}).roles || ['User'];
let idTokenClaims = context.idToken || {};
idTokenClaims[`nussknacker:roles`] = assignedRoles;
context.idToken = idTokenClaims;
callback(null, user, context);}
You can find more configurations options at Auth0's documentation on applications and APIs
In Nussknacker's configuration file add the following authentication
section:
authentication: {
method: "Oidc"
issuer: "https://<the value of Applications -> Application Name -> Settings -> Basic Information -> Domain>"
clientSecret: "<the value of Applications -> Application Name -> Settings -> Basic Information -> Client Secret>"
clientId: "<the value of Applications -> Application Name -> Settings -> Basic Information -> Client Identifier>"
audience: "<the value of APIs -> API Name -> Settings -> General Settings -> Identifier>"
rolesClaim: "nussknacker:roles"
usersFile: "conf/users.conf"
}
The role names in the usersFile
should match the roles defined in the Auth0 tenant.
OAuth2 security module
Generic configuration
authentication: {
method: ${?AUTHENTICATION_METHOD} (default: "BasicAuth")
clientSecret: ${?OAUTH2_CLIENT_SECRET}
clientId: ${?OAUTH2_CLIENT_ID}
authorizeUri: ${?OAUTH2_AUTHORIZE_URI}
redirectUri: ${?OAUTH2_REDIRECT_URI}
accessTokenUri: ${?OAUTH2_ACCESS_TOKEN_URI}
profileUri: ${?OAUTH2_PROFILE_URI}
profileFormat: ${?OAUTH2_PROFILE_FORMAT}
implicitGrantEnabled: ${?OAUTH2_IMPLICIT_GRANT_ENABLED}
jwt {
accessTokenIsJwt: ${?OAUTH2_ACCESS_TOKEN_IS_JWT} (default: false)
userinfoFromIdToken: ${?OAUTH2_USERINFO_FROM_ID_TOKEN} (default: false)
audience: ${?OAUTH2_JWT_AUDIENCE}
publicKey: ${?OAUTH2_JWT_AUTH_SERVER_PUBLIC_KEY}
publicKeyFile: ${?OAUTH2_JWT_AUTH_SERVER_PUBLIC_KEY_FILE}
certificate: ${?OAUTH2_JWT_AUTH_SERVER_CERTIFICATE}
certificateFile: ${?OAUTH2_JWT_AUTH_SERVER_CERTIFICATE_FILE}
idTokenNonceVerificationRequired: ${?OAUTH2_JWT_ID_TOKEN_NONCE_VERIFICATION_REQUIRED}
}
accessTokenParams {
grant_type: ${?OAUTH2_GRANT_TYPE}
}
accessTokenRequestContentType: "application/json" (default)
authorizeParams {
response_type: ${?OAUTH2_RESPONSE_TYPE}
scope: ${?OAUTH2_SCOPE}
audience: ${?OAUTH2_AUDIENCE}
}
headers {
Accept: ${?AUTHENTICATION_HEADERS_ACCEPT}
}
usersFile: ${?AUTHENTICATION_USERS_FILE}
}
When method
is set to OAuth2
, the following fields are mandatory: clientSecret
, clientId
, authorizeUri
, redirectUri
, accessTokenUri
, profileUri
, profileFormat
, implicitGrantEnabled
, usersFile
.
For the profileFormat
one of oidc
or github
is supported by default.
Subconfigs accessTokenParams
, accessTokenRequestContentType
, authorizeParams
, headers
are optional and every
field from any of the subconfigs is optional and could be provided separately.
By default, access token request is sent using application/json
content type, to change it (e.g.
to application/x-www-form-urlencoded
) use accessTokenRequestContentType
config.
Subconfig jwt
is also optional. However, if it is present and enabled
is set to
true, the audience
and one of the publicKey
, publicKeyFile
, certificate
, certificateFile
,
fields have to be provided.
Access tokens are introspected only once and then stored in a cache for the expiration time.
When using JWT, an incoming access token is validated against its signature and aud
(audience) claim.
Otherwise, a token is considered valid when it allows a profileUri
(e.g. /userinfo
) query.
No token refreshing nor revoking is implemented.
Remarks:
-
Setting
jwt.idTokenNonceVerificationRequired
totrue
has an effect only ifimplicitGrantEnabled
is also set totrue
. -
For security reasons, if implicit flow used,
implicitGrantEnabled
andjwt.idTokenNonceVerificationRequired
should be set totrue
. FurthermoreauthorizeParams.response_type
should be set to"access_token id_token"
(not only"access_token"
). Then a receivedaccess_token
format has to be JWT (likeid_token
).Of course, this does not apply when
authorizeParams.response_type
is set tocode
(code flow used). -
Provided
jwt
is enabled, the backend first checks whether a user profile can be obtained from theaccess_token
, secondly it tries to obtain the profile from a request sent toauthorizeUri
.
OAuth2 security module - GitHub example with code flow
Configuration in following format:
authentication: {
method: "OAuth2"
clientSecret: ""
clientId: ""
authorizeUri: "https://github.com/login/oauth/authorize"
redirectUri: "http://localhost:3000"
accessTokenUri: "https://github.com/login/oauth/access_token"
profileUri: "https://api.github.com/user"
profileFormat: "github"
implicitGrantEnabled: false
accessTokenParams {
grant_type: "authorization_code"
}
headers {
Accept: "application/json"
}
authorizeParams {
response_type: "code"
}
usersFile: "./src/test/resources/oauth2-users.conf"
}
Users file in following format:
users: [ //Special settings by user email
{
identity: "some@email.com"
roles: ["Admin"]
}
]
rules: [
{
role: "Admin"
isAdmin: true
},
{
role: "User" //this is default role for all users
permissions: ["Read", "Write", "Deploy"]
categories: ["Defautl", "FraudDetection"]
}
]
UI customization options
Process Toolbar Configuration
Toolbars and buttons at process window are configurable, you can configure params:
- uuid - optional uuid identifier which determines unique code for FE localstorage cache, default: null - we generate uuid from hashcode config
- topLeft - optional top left panel, default: empty list
- bottomLeft - optional bottom left panel, default: empty list
- topRight - optional top right panel, default: empty list
- bottomRight - optional bottom right panel, default: empty list
Example configuration:
processToolbarConfig {
defaultConfig {
topLeft: [
{ type: "tips-panel" }
]
topRight: [
{
type: "process-info-panel"
buttons: [
{ type: "process-save", disabled: { subprocess: false, archived: true, type: "oneof" } }
{ type: "custom-link", title: "Metrics for $processName", url: "/metrics/$processId" }
]
}
]
}
}
We can also create special configuration for each category by:
processToolbarConfig {
categoryConfig {
"CategoryName" {
id: "58f1acff-d864-4d66-9f86-0fa7319f7043"
topLeft: [
{ type: "creator-panel", hidden: {subprocess: true, archived: false, type: "allof"} }
]
}
}
}
Toolbar Panel Conditioning
Configuration allow us to:
- hiding panels
- hiding or disabling buttons
Each of this function can be configured by condition expression where we can use three parameters:
subprocess: boolean
- if true then condition match only subprocess, by default ignoredarchived: boolean
- if true then condition match only archived, by default ignoredtype: allof / oneof
- information about that checked will be only one condition or all conditions
Toolbar Panel Templating
Configuration allows to templating params like:
name
- available only on Buttonstitle
- available on Panels and Buttonsurl
- available only on CustomLink and CustomAction buttonsicon
- available only on Buttons
Right now we allow to template two elements:
- scenario id -
$processId
- scenario name -
$processName
Example usage:
title: "Metrics for $processName"
name: "deploy $processName"
url: "/metrics/$processId"
icon: "/assets/process-icon-$processId"
Default Process Panel Configuration
processToolbarConfig {
defaultConfig {
topLeft: [
{ type: "tips-panel" }
{ type: "creator-panel", hidden: { archived: true } }
{ type: "versions-panel" }
{ type: "comments-panel" }
{ type: "attachments-panel" }
]
topRight: [
{
type: "process-info-panel"
buttons: [
{ type: "process-save", title: "Save changes", disabled: { archived: true } }
{ type: "process-deploy", disabled: { subprocess: true, archived: true, type: "oneof" } }
{ type: "process-cancel", disabled: { subprocess: true, archived: true, type: "oneof" } }
{ type: "custom-link", name: "metrics", icon: "/assets/buttons/metrics.svg", url: "/metrics/$processName", disabled: { subprocess: true } }
]
}
{
id: "view-panel"
type: "buttons-panel"
title: "view"
buttons: [
{ type: "view-zoom-in" }
{ type: "view-zoom-out" }
{ type: "view-reset" }
]
}
{
id: "edit-panel"
type: "buttons-panel"
title: "edit"
hidden: { archived: true }
buttonsVariant: "small"
buttons: [
{ type: "edit-undo" }
{ type: "edit-redo" }
{ type: "edit-copy" }
{ type: "edit-paste" }
{ type: "edit-delete" }
{ type: "edit-layout" }
]
}
{
id: "process-panel"
type: "buttons-panel"
title: "process"
buttons: [
{ type: "process-properties", hidden: { subprocess: true } }
{ type: "process-compare" }
{ type: "process-migrate", disabled: { archived: true } }
{ type: "process-import", disabled: { archived: true } }
{ type: "process-json" }
{ type: "process-pdf" }
{ type: "process-archive", hidden: { archived: true } }
{ type: "process-unarchive", hidden: { archived: false } }
]
}
{
id: "test-panel"
type: "buttons-panel"
title: "test"
hidden: { subprocess: true }
buttons: [
{ type: "test-from-file", disabled: { archived: true } }
{ type: "test-generate", disabled: { archived: true } }
{ type: "test-counts" }
{ type: "test-hide" }
]
}
{
id: "group-panel"
type: "buttons-panel"
title: "group"
hidden: { archived: true }
buttons: [
{ type: "group" }
{ type: "ungroup" }
]
}
]
}
}
Main menu configuration
Tabs (in main menu bar, such as Scenarios etc.) can be configured in the following way:
tabs: ${tabs} [
{
id: "kibana",
title: "Kibana",
url: "https://myKibana.org/kibana",
type: "IFrame",
requiredPermission: "AdminTab"
}
]
By default, only Scenarios
tab is configured.
Parameter name | Type | Description |
---|---|---|
id | string | Unique identifier |
title | string | Title appearing in UI |
type | IFrame/Local/Remote/Url | Type of tab (see below for explanation) |
url | string | URL of the tab |
requiredPermission | string | Optional parameter, name of Global Permission |
The types of tabs can be as follows (see dev-application.conf
for some examples):
- IFrame - contents of the url parameter will be embedded as IFrame
- Local - redirect to Designer page (
/admin
,/signals
,/processes
etc., see code for other options) - Remote - module federation can be used to embed external tabs, url should be in form:
{module}/{path}@{host}/{remoteEntry}.js
- Url - redirect to external page/url
Environment configuration
Nussknacker installation may consist of more than one environment. Typical example is a configuration, where you have:
- environment for testing scenarios, which mirrors production data (e.g. via Kafka mirror-maker), but contains only scenarios that are currently worked on
- production environment
You can configure secondaryEnvironment
to allow for
- easy migration of scenarios
- comparing scenarios between environments
- testing (currently only via REST API) if all scenarios from secondary environment are valid with model configuration from this environment (useful for testing configuration etc.) Currently, you can only configure secondary environment if it uses BASIC authentication - technical user is needed to access REST API.
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
environment | Medium | string | Used mainly for metrics configuration. Please note: it has to be consistent with tag configuration of metrics | |
environmentAlert.content | Low | string | Human readable name of environment, to display in UI | |
environmentAlert.cssClass | Low | indicator-green / indicator-blue / indicator-yellow / indicator-red | Color of environment indicator | |
secondaryEnvironment.remoteConfig.uri | Medium | string | URL of Nussknacker REST API e.g. http://secondary.host:8080/api | |
secondaryEnvironment.remoteConfig.batchSize | Low | int | 10 | For testing compatibility we have to load all scenarios, we do it in batches to optimize |
secondaryEnvironment.user | Medium | string | User that should be used for migration/comparison | |
secondaryEnvironment.password | Medium | string | Password of the user that should be used for migration/comparison | |
secondaryEnvironment.targetEnvironmentId | Low | string | Name of the secondary environment (used mainly for messages for user) |
Testing
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
testDataSettings.maxSampleCount | Medium | string | 20 | Limits number of samples for tests from file |
testDataSettings.testDataMaxBytes | Low | string | 200000 | Limits size of test input for tests from file |
testDataSettings.resultsMaxBytes | Low | string | 50000000 | Limits size of returned test data for tests from file |
Other configuration options
Parameter name | Importance | Type | Default value | Description |
---|---|---|---|---|
attachments.maxSizeInBytes | Medium | long | 10485760 | Limits max size of scenario attachment, by default to 10mb |
analytics.engine | Low | Matomo | Currently only available analytics engine is Matomo | |
analytics.url | Low | string | URL of Matomo server | |
analytics.siteId | Low | string | Site id | |
intervalTimeSettings.processes | Low | int | 20000 | How often frontend reloads scenario list |
intervalTimeSettings.healthCheck | Low | int | 30000 | How often frontend reloads checks scenarios states |
developmentMode | Medium | boolean | false | For development mode we disable some security features like CORS. Don't use in production |
Scenario type, categories
Every scenario has to belong to a group called category
. For example, in one Nussknacker installation you can have
scenarios detecting frauds, and those implementing marketing campaigns. Category configuration looks like this:
categoriesConfig: {
"marketing": "streaming",
"fraud": "streaming",
}
For each category you have to define its scenario type (streaming
in examples above). Scenario type configuration consists of two parts:
deploymentConfig
- deployment manager configurationmodelConfig
- model configuration
In Nussknacker distribution there are preconfigured scenario types:
streaming
- using Flink Deployment Manager providing both stateful and stateless streaming componentsstreaming-lite-embedded
- using embedded Streaming-Lite Deployment Manager providing only stateless streaming componentsrequest-response-embedded
- use embedded Request-Response Deployment Manager, scenario logic is exposed as REST API, on additional HTTP port at Designer
And one Default
category using streaming
by default (can be configured via DEFAULT_SCENARIO_TYPE
environment variable)
See example from development config for more complex examples.