Kerberos Authentication Plugin

If you are using Kerberos to secure your network environment, the Kerberos authentication plugin can be used to secure a Solr cluster.

This allows Solr to use a Kerberos service principal and keytab file to authenticate with ZooKeeper and between nodes of the Solr cluster (if applicable). Users of the Admin UI and all clients (such as SolrJ) would also need to have a valid ticket before being able to use the UI or send requests to Solr.

Support for Kerberos authentication is available in SolrCloud, user-managed, or single-node installations.

If you are using Solr with a Hadoop cluster secured with Kerberos and intend to store your Solr indexes in HDFS, also see the section Solr on HDFS for additional steps to configure Solr for that purpose. The instructions on this page apply only to scenarios where Solr will be secured with Kerberos. If you only need to store your indexes in a Kerberized HDFS system, please see the Running Solr on HDFS section.

How Solr Works With Kerberos

When setting up Solr to use Kerberos, configurations are put in place for Solr to use a service principal, or a Kerberos username, which is registered with the Key Distribution Center (KDC) to authenticate requests. The configurations define the service principal name and the location of the keytab file that contains the credentials.

As with all authentication plugins, Kerberos authentication configuration is stored in security.json. This file is discussed in the section Configuring security.json.

Service Principals and Keytab Files

Each Solr node must have a service principal registered with the Key Distribution Center (KDC). The Kerberos plugin uses SPNego to negotiate authentication.

Using HTTP/host1@YOUR-DOMAIN.ORG as an example of a service principal:

  • HTTP indicates the type of requests which this service principal will be used to authenticate. The HTTP/ in the service principal is a must for SPNego to work with requests to Solr over HTTP.

  • host1 is the host name of the machine hosting the Solr node.

  • YOUR-DOMAIN.ORG is the organization-wide Kerberos realm.

Multiple Solr nodes on the same host may have the same service principal, since the host name is common to them all.

Along with the service principal, each Solr node needs a keytab file which should contain the credentials of the service principal used. A keytab file contains encrypted credentials to support passwordless logins while obtaining Kerberos tickets from the KDC. For each Solr node, the keytab file should be kept in a secure location and not shared with users of the cluster.

Since a Solr cluster requires internode communication, each node must also be able to make Kerberos enabled requests to other nodes. By default, Solr uses the same service principal and keytab as a 'client principal' for internode communication. You may configure a distinct client principal explicitly, but doing so is not recommended and is not covered in the examples below.

Kerberized ZooKeeper

When setting up a kerberized SolrCloud cluster, it is recommended to enable Kerberos security for ZooKeeper as well.

In such a setup, the client principal used to authenticate requests with ZooKeeper can be shared for internode communication as well. This has the benefit of not needing to renew the ticket granting tickets (TGTs) separately, since the ZooKeeper client used by Solr takes care of this. To achieve this, a single JAAS configuration (with the app name as Client) can be used for the Kerberos plugin as well as for the ZooKeeper client.

See the ZooKeeper Configuration section below for an example of starting ZooKeeper in Kerberos mode.

Browser Configuration

In order for your browser to access the Solr Admin UI after enabling Kerberos authentication, it must be able to negotiate with the Kerberos authenticator service to allow you access. Each browser supports this differently, and some (like Chrome) do not support it at all. If you see 401 errors when trying to access the Solr Admin UI after enabling Kerberos authentication, it’s likely your browser has not been configured properly to know how or where to negotiate the authentication request.

Detailed information on how to set up your browser is beyond the scope of this documentation; please see your system administrators for Kerberos for details on how to configure your browser.

Kerberos Authentication Configuration

Consult Your Kerberos Admins!

Before attempting to configure Solr to use Kerberos authentication, please review each step outlined below and consult with your local Kerberos administrators on each detail to be sure you know the correct values for each parameter. Small errors can cause Solr to not start or not function properly, and are notoriously difficult to diagnose.

Configuration of the Kerberos plugin has several parts:

  • Create service principals and keytab files

  • ZooKeeper configuration

  • Create or update /security.json

  • Define jaas-client.conf

  • Solr startup parameters

We’ll walk through each of these steps below.

Using Hostnames

To use host names instead of IP addresses, use the SOLR_HOST configuration in bin/solr.in.sh or pass a -Dhost=<hostname> system parameter during Solr startup. This guide uses IP addresses. If you specify a hostname, replace all the IP addresses in the guide with the Solr hostname as appropriate.

Get Service Principals and Keytabs

Before configuring Solr, make sure you have a Kerberos service principal for each Solr host and ZooKeeper (if ZooKeeper has not already been configured) available in the KDC server, and generate a keytab file as shown below.

This example assumes the hostname is 192.168.0.107 and your home directory is /home/foo/. This example should be modified for your own environment.

root@kdc:/# kadmin.local
Authenticating as principal foo/admin@EXAMPLE.COM with password.

kadmin.local:  addprinc HTTP/192.168.0.107
WARNING: no policy specified for HTTP/192.168.0.107@EXAMPLE.COM; defaulting to no policy
Enter password for principal "HTTP/192.168.0.107@EXAMPLE.COM":
Re-enter password for principal "HTTP/192.168.0.107@EXAMPLE.COM":
Principal "HTTP/192.168.0.107@EXAMPLE.COM" created.

kadmin.local:  ktadd -k /tmp/107.keytab HTTP/192.168.0.107
Entry for principal HTTP/192.168.0.107 with kvno 2, encryption type aes256-cts-hmac-sha1-96 added to keytab WRFILE:/tmp/107.keytab.
Entry for principal HTTP/192.168.0.107 with kvno 2, encryption type arcfour-hmac added to keytab WRFILE:/tmp/107.keytab.
Entry for principal HTTP/192.168.0.107 with kvno 2, encryption type des3-cbc-sha1 added to keytab WRFILE:/tmp/108.keytab.
Entry for principal HTTP/192.168.0.107 with kvno 2, encryption type des-cbc-crc added to keytab WRFILE:/tmp/107.keytab.

kadmin.local:  quit

Copy the keytab file from the KDC server’s /tmp/107.keytab location to the Solr host at /keytabs/107.keytab. Repeat this step for each Solr node.

You might need to take similar steps to create a ZooKeeper service principal and keytab if it has not already been set up. In that case, the example below shows a different service principal for ZooKeeper, so the above might be repeated with zookeeper/host1 as the service principal for one of the nodes

ZooKeeper Configuration

If you are using a ZooKeeper that has already been configured to use Kerberos, you can skip the ZooKeeper-related steps shown here.

Since ZooKeeper manages the communication between nodes in a SolrCloud cluster, it must also be able to authenticate with each node of the cluster. Configuration requires setting up a service principal for ZooKeeper, defining a JAAS configuration file and instructing ZooKeeper to use both of those items.

The first step is to create a file java.env in ZooKeeper’s conf directory and add the following to it, as in this example:

export JVMFLAGS="-Djava.security.auth.login.config=/etc/zookeeper/conf/jaas-client.conf"

The JAAS configuration file should contain the following parameters. Be sure to change the principal and keyTab path as appropriate. The file must be located in the path defined in the step above, with the filename specified.

Server {
 com.sun.security.auth.module.Krb5LoginModule required
  useKeyTab=true
  keyTab="/keytabs/zkhost1.keytab"
  storeKey=true
  doNotPrompt=true
  useTicketCache=false
  debug=true
  principal="zookeeper/host1@EXAMPLE.COM";
};

Finally, add the following lines to the ZooKeeper configuration file zoo.cfg:

authProvider.1=org.apache.zookeeper.server.auth.SASLAuthenticationProvider
jaasLoginRenew=3600000

Once all of the pieces are in place, start ZooKeeper with the following parameter pointing to the JAAS configuration file:

$ bin/zkServer.sh start -Djava.security.auth.login.config=/etc/zookeeper/conf/jaas-client.conf

Create security.json

Create the file security.json with the contents:

{"authentication": {"class": "solr.KerberosPlugin"}}

Then use the bin/solr zk command to upload the file:

$ bin/solr zk cp ./security.json zk:security.json -z localhost:2181

If you are using Solr in a single-node installation, you need to create the security.json file and put it in your $SOLR_HOME directory.

If you already have a /security.json file in ZooKeeper, download the file, add or modify the authentication section and upload it back to ZooKeeper using the ZooKeeper Utilities available in Solr.

Define a JAAS Configuration File

The JAAS configuration file defines the properties to use for authentication, such as the service principal and the location of the keytab file. Other properties can also be set to ensure ticket caching and other features.

The following example can be copied and modified slightly for your environment. The location of the file can be anywhere on the server, but it will be referenced when starting Solr so it must be readable on the filesystem. The JAAS file may contain multiple sections for different users, but each section must have a unique name so it can be uniquely referenced in each application.

In the below example, we have created a JAAS configuration file with the name and path of /home/foo/jaas-client.conf. We will use this name and path when we define the Solr start parameters in the next section. Note that the client principal here is the same as the service principal. This will be used to authenticate internode requests and requests to ZooKeeper. Make sure to use the correct principal hostname and the keyTab file path.

Client {
  com.sun.security.auth.module.Krb5LoginModule required
  useKeyTab=true
  keyTab="/keytabs/107.keytab"
  storeKey=true
  useTicketCache=true
  debug=true
  principal="HTTP/192.168.0.107@EXAMPLE.COM";
};

The first line of this file defines the section name, which will be used with the solr.kerberos.jaas.appname parameter, defined below.

The main properties we are concerned with are the keyTab and principal properties, but there are others which may be required for your environment. The javadocs for the Krb5LoginModule (the class that’s being used and is called in the second line above) provide a good outline of the available properties, but for reference the ones in use in the above example are explained here:

  • useKeyTab: this boolean property defines if we should use a keytab file (true, in this case).

  • keyTab: the location and name of the keytab file for the principal this section of the JAAS configuration file is for. The path should be enclosed in double-quotes.

  • storeKey: this boolean property allows the key to be stored in the private credentials of the user.

  • useTicketCache: this boolean property allows the ticket to be obtained from the ticket cache.

  • debug: this boolean property will output debug messages for help in troubleshooting.

  • principal: the name of the service principal to be used.

Solr Startup Parameters

While starting up Solr, the following host-specific parameters need to be passed. These parameters can be passed at the command line with the bin/solr start command (see Solr Control Script Reference for details on how to pass system parameters) or defined in bin/solr.in.sh or bin/solr.in.cmd as appropriate for your operating system.

solr.kerberos.name.rules

Optional

Default: DEFAULT

Used to map Kerberos principals to short names. Example of a name rule: RULE:[1:$1@$0](.*EXAMPLE.COM)s/@.*//.

solr.kerberos.name.rules.mechanism

Optional

Default: hadoop

The mechanism used to map Kerberos principals to short names. This can be either hadoop or mit.

solr.kerberos.cookie.domain

Required

Default: none

Used to issue cookies and should have the hostname of the Solr node.

solr.kerberos.cookie.portaware

Optional

Default: false

When set to true, cookies are differentiated based on host and port, as opposed to standard cookies which are not port aware. This should be set if more than one Solr node is hosted on the same host.

solr.kerberos.principal

Required

Default: none

The service principal.

solr.kerberos.keytab

Required

Default: none

Keytab file path containing service principal credentials

solr.kerberos.jaas.appname

Optional

Default: Client

The app name (section name) within the JAAS configuration file which is required for internode communication. The default is used for ZooKeeper authentication as well. If different users are used for ZooKeeper and Solr, they will need to have separate sections in the JAAS configuration file.

java.security.auth.login.config

Required

Default: none

Path to the JAAS configuration file for configuring a Solr client for internode communication.

Here is an example that could be added to bin/solr.in.sh. Make sure to change this example to use the right hostname and the keytab file path.

SOLR_AUTH_TYPE="kerberos"
SOLR_AUTHENTICATION_OPTS="-Djava.security.auth.login.config=/home/foo/jaas-client.conf -Dsolr.kerberos.cookie.domain=192.168.0.107 -Dsolr.kerberos.cookie.portaware=true -Dsolr.kerberos.principal=HTTP/192.168.0.107@EXAMPLE.COM -Dsolr.kerberos.keytab=/keytabs/107.keytab"
KDC with AES-256 encryption

If your KDC uses AES-256 encryption, you need to add the Java Cryptography Extension (JCE) Unlimited Strength Jurisdiction Policy Files to your JRE before a Kerberized Solr can interact with the KDC.

You will know this when you see an error like this in your Solr logs: "KrbException: Encryption type AES256 CTS mode with HMAC SHA1-96 is not supported/enabled".

For Java 1.8, this is available here: http://www.oracle.com/technetwork/java/javase/downloads/jce8-download-2133166.html.

Replace the local_policy.jar present in JAVA_HOME/jre/lib/security/ with the new local_policy.jar from the downloaded package and restart the Solr node.

Using Delegation Tokens

The Kerberos plugin can be configured to use delegation tokens, which allow an application to reuse the authentication of an end-user or another application.

There are a few use cases for Solr where this might be helpful:

  • Using distributed clients (such as MapReduce) where each client may not have access to the user’s credentials.

  • When load on the Kerberos server is high. Delegation tokens can reduce the load because they do not access the server after the first request.

  • If requests or permissions need to be delegated to another user.

To enable delegation tokens, several parameters must be defined. These parameters can be passed at the command line with the bin/solr start command Solr Control Script Reference) or defined in bin/solr.in.sh or bin/solr.in.cmd as appropriate for your operating system.

solr.kerberos.delegation.token.enabled

Optional

Default: false

Set to true to enable delegation tokens. This parameter is required if you want to enable tokens.

solr.kerberos.delegation.token.kind

Optional

Default: solr-dt

The type of delegation tokens. The only option available at this time is the default.

solr.kerberos.delegation.token.validity

Optional

Default: 36000

Time, in seconds, for which delegation tokens are valid.

solr.kerberos.delegation.token.signer.secret.provider

Optional

Default: zookeeper

Where delegation token information is stored internally. The default is zookeeper which must be the location for delegation tokens to work across Solr servers (when running in SolrCloud mode). No other option is available at this time.

solr.kerberos.delegation.token.signer.secret.provider.zookeper.path

Optional

Default: none

The ZooKeeper path where the secret provider information is stored. This is in the form of the path + /security/token. The path can include the chroot or the chroot can be omitted if you are not using it. This example includes the chroot: server1:9983,server2:9983,server3:9983/solr/security/token.

solr.kerberos.delegation.token.secret.manager.znode.working.path

Optional

Default: none

The ZooKeeper path where token information is stored. This is in the form of the path + /security/zkdtsm. The path can include the chroot or the chroot can be omitted if you are not using it. This example includes the chroot: server1:9983,server2:9983,server3:9983/solr/security/zkdtsm.

Start Solr

Once the configuration is complete, you can start Solr with the bin/solr script, as in the example below, which is for users in SolrCloud mode only.

This example assumes you modified bin/solr.in.sh or bin/solr.in.cmd, with the proper values, but if you did not, you would pass the system parameters along with the start command. Note you also need to customize the -z property as appropriate for the location of your ZooKeeper nodes.

$ bin/solr -c -z server1:2181,server2:2181,server3:2181/solr
If you have defined ZK_HOST in solr.in.sh/solr.in.cmd (see Updating Solr Include Files) you can omit -z <zk host string> from the above command.

Test the Configuration

  1. Do a kinit with your username. For example, kinit user@EXAMPLE.COM.

  2. Try to access Solr using curl. You should get a successful response.

    $ curl --negotiate -u : "http://192.168.0.107:8983/solr/"

Using SolrJ with a Kerberized Solr

To use Kerberos authentication in a SolrJ application, you need the following two lines before you create a SolrClient:

System.setProperty("java.security.auth.login.config", "/home/foo/jaas-client.conf");
HttpClientUtil.setHttpClientBuilder(new Krb5HttpClientBuilder().getBuilder());

You need to specify a Kerberos service principal for the client and a corresponding keytab in the JAAS client configuration file above. This principal should be different from the service principal we created for Solr.

Here’s an example:

SolrJClient {
  com.sun.security.auth.module.Krb5LoginModule required
  useKeyTab=true
  keyTab="/keytabs/foo.keytab"
  storeKey=true
  useTicketCache=true
  debug=true
  principal="solrclient@EXAMPLE.COM";
};

Delegation Tokens with SolrJ

Delegation tokens are also supported with SolrJ, in the following ways:

  • DelegationTokenRequest and DelegationTokenResponse can be used to get, cancel, and renew delegation tokens.

  • HttpSolrClient.Builder includes a withKerberosDelegationToken function for creating an HttpSolrClient that uses a delegation token to authenticate.

Sample code to get a delegation token:

private String getDelegationToken(final String renewer, final String user, HttpSolrClient solrClient) throws Exception {
    DelegationTokenRequest.Get get = new DelegationTokenRequest.Get(renewer) {
      @Override
      public SolrParams getParams() {
        ModifiableSolrParams params = new ModifiableSolrParams(super.getParams());
        params.set("user", user);
        return params;
      }
    };
    DelegationTokenResponse.Get getResponse = get.process(solrClient);
    return getResponse.getDelegationToken();
  }

To create a HttpSolrClient that uses delegation tokens:

HttpSolrClient client = new HttpSolrClient.Builder("http://localhost:8983/solr").withKerberosDelegationToken(token).build();

To create a CloudSolrClient that uses delegation tokens:

CloudSolrClient client = new CloudSolrClient.Builder(Collections.singletonList("localhost:2181"),Optional.empty())
                .withLBHttpSolrClientBuilder(new LBHttpSolrClient.Builder()
                    .withResponseParser(client.getParser())
                    .withHttpSolrClientBuilder(
                        new HttpSolrClient.Builder()
                            .withKerberosDelegationToken(token)
                    ))
                        .build();

Hadoop’s delegation token responses are in JSON map format. A response parser for that is available in DelegationTokenResponse. Other response parsers may not work well with Hadoop responses.