[[_vault-administration]] == Using Vault to Obtain Secrets Several fields in the administration support obtaining the value of a secret from an external vault. To obtain a secret from a vault instead of entering it directly, enter the following specially crafted string into the appropriate field: `**${vault.**_key_**}**` where you replace the `_key_` with the name of the secret as recognized by the vault. In order to prevent secrets from leaking across realms, implementations may combine the realm name with the `_key_` obtained from the vault expression. This means that the `_key_` won't directly map to an entry in the vault, but rather be used to create the final entry name according to the algorithm used to combine it with the realm name. Currently, the secret can be obtained from the vault in the following fields: SMTP password:: In realm <<_email,SMTP settings>> LDAP bind credential:: In <<_ldap,LDAP settings>> of LDAP-based user federation. OIDC identity provider secret:: In _Client Secret_ inside identity provider <<_identity_broker_oidc,OpenID Connect Config>> To use a vault, a vault provider must be registered within {project_name}. It is possible to either use a built-in provider described below or implement your own provider. See the link:{developerguide_link}[{developerguide_name}] for more information. NOTE: There is at most one vault provider active per {project_name} instance at any given time, and the vault provider in each instance within the cluster has to be configured consistently. === Kubernetes / OpenShift Files Plaintext Vault Provider {project_name} supports vault implementation for https://kubernetes.io/docs/concepts/configuration/secret/[Kubernetes secrets]. These secrets can be mounted as data volumes, and they appear as a directory with a flat file structure, where each secret is represented by a file whose name is the secret name, and contents of that file is the secret value. The files within this directory have to be named as secret name prefixed by realm name and an underscore. All underscores within the secret name or the realm name have to be doubled in the file name. For example, for a field within a realm called `sso_realm`, a reference to a secret with name `secret-name` would be written as `${vault.secret-name}`, and the file name looked up would be `sso+++__+++realm+++_+++secret-name` (note the underscore doubled in realm name). To use this type of secret store, you have to declare the `files-plaintext` vault provider in standalone.xml, and set its parameter for the directory that contains the mounted volume. The following example shows the `files-plaintext` provider with the directory where vault files are searched for set to `standalone/configuration/vault` relative to {project_name} base directory: [source, xml] ---- files-plaintext ---- Here is the equivalent configuration using CLI commands: [source,bash] ---- /subsystem=keycloak-server/spi=vault/:add /subsystem=keycloak-server/spi=vault/provider=files-plaintext/:add(enabled=true,properties={dir => "${jboss.home.dir}/standalone/configuration/vault"}) ---- === Elytron Credential Store Vault Provider {project_name} also provides support for reading secrets stored in an Elytron credential store. The `elytron-cs-keystore` vault provider is capable of retrieving secrets from the keystore-based implementation of the credential store, which is also the default implementation provided by Elytron. This credential store is backed by a keystore (`JCEKS` is the default format, but it is possible to use other formats such as `PKCS12`) and users can create and manage the store contents using either the `elytron` subsystem in WildFly/JBoss EAP, or using the `elytron-tool.sh` script. To use this provider, you have to declare the `elytron-cs-keystore` in the `keycloak-server` subsystem and set the location and master secret of the keystore that was created by Elytron. An example of the minimal configuration for the provider follows: [source, xml] ---- elytron-cs-keystore ---- If the underlying keystore has a format other than `JCEKS`, this format has to be informed using the `keyStoreType`: [source, xml] ---- elytron-cs-keystore ---- For the secret, the `elytron-cs-keystore` provider supports both clear-text values (as shown above) and also values that were masked using the `elytron-tool.sh` script: [source, xml] ---- ... ... ---- For more detailed information on how to create/manage elytron credential stores, as well as how to mask keystore secrets, please refer to the Elytron documentation. NOTE: The `elytron-cs-keystore` vault provider has been implemented as a WildFly extension and as such is only available if the {project_name} server runs on WildFly/JBoss EAP. === Key Resolvers All built-in providers support the configuration of one or more key resolvers. A key resolver essentially implements the algorithm or strategy for combining the realm name with the key (as obtained from the `${vault.key}` expression} into the final entry name that will be used to retrieve the secret from the vault. The `keyResolvers` property is used to configure the resolvers that are to be used by the provider. The value is a comma-separated list of resolver names. An example of configuration for the `files-plaintext` provider follows: [source, xml] ---- files-plaintext ---- The resolvers are executed in the same order that they are declared in the configuration. For each resolver, the final entry name produced by the resolver that combines the realm with the vault key is used to search for the secret in the vault. If a secret is found, it is immediately returned. If not, the next resolver is used and this continues until a non-empty secret is found or all resolvers have been tried, in which case an empty secret is returned. In the example above, first the `REALM_UNDERSCORE_KEY` resolver is used. If an entry is found in the vault with the name it produces, it is returned. If not, then the `KEY_ONLY` resolver is used. Again, if an entry is found in the vault with the name it produces, it is returned. If not, an empty secret is returned since there are no more resolvers to be used. A list of the currently available resolvers follows: * `KEY_ONLY`: the realm name is ignored and the key from the vault expression is used as is. * `REALM_UNDERSCORE_KEY`: the realm and key are combined using an underscore `_` character. Occurrences of underscore in either the realm or key are escaped by another underscore character. So if the realm is called `master_realm` and the key is `smtp_key`, the combined key will be `master+++__+++realm_smtp+++__+++key`. * `REALM_FILESEPARATOR_KEY`: the realm and key are combined using the platform file separator character. This is useful in situations where the keys are grouped by realm using a directory structure. ifeval::[{project_community}==true] * `FACTORY_PROVIDED`: the realm and key are combined using the `VaultKeyResolver` that is provided by the vault provider factory, allowing the creation of a custom key resolver by extending an existing factory and implementing the `getFactoryResolver` method. endif::[] If no resolver is configured for the built-in providers, the `REALM_UNDERSCORE_KEY` is selected by default. ifeval::[{project_community}==true] The `FACTORY_PROVIDED` resolver provides a hook that can be used to implement a custom resolver by extending the provider factory of choice and overriding the `getFactoryResolver` method so it returns the custom resolver. For example, if you want to use the `elytron-cs-keystore` provider but none of the built-in resolvers match the format used in your keystore, you can extend the `ElytronCSKeystoreProvider and implement the getFactoryResolver method: [source,java] ---- public class CustomElytronProviderFactory extends ElytronCSKeyStoreProviderFactory { ... @Override protected VaultKeyResolver getFactoryResolver() { return (realm, key) -> realm + "###" + key; } @Override public String getId() { return "custom-elytron-cs-keystore; } ... } ---- The custom factory returns a key resolver that combines the realm and key with a triple `#` character. So an entry would look like `master_realm###smtp_key`, for example. This factory must then be installed just like any custom provider. Note that the custom factory must override both the `getFactoryResolver` and `getId` methods. The second method is needed so that we can properly configure the custom factory in {project_name}. To install and use the above custom provider the configuration would look something like this: [source, xml] ---- custom-elytron-cs-keystore ---- The configuration above tells {project_name} to setup the custom Elytron provider and use the key resolver that is created by the custom factory. endif::[]