Hazelcast Python Client Kerberos Authentication Support

Requirements

• Hazelcast Enterprise 4.1 and up,
• Hazelcast Python Client 4.2.1 and up,
• Python 3.5 or better,
• Linux with a recent Kernel (tested on Debian Stretch and Buster),
• Kerberos 5 shared library (libkrb5-dev package on Debian/Ubuntu systems),
• C compiler is required to build the python-gssapi dependency. You may opt to use the system provided package (python3-gssapi on Debian/Ubuntu systems),

Install

pip install -U hazelcast-kerberos

Usage

Enabling Kerberos authentication for Hazelcast involves server and client configuration.

Client Configuration

On the client side, a Kerberos token provider is created and passed to Hazelcast Python Client. The token provider authenticates to KDC using the given credentials, receives and caches the Kerberos ticket and retrieves a token. The token is passed to the server-side by Hazelcast Python Client for authenticating to the server.

Using a Cached Ticket

If a Kerberos ticket was already cached, probably using the kinit command, then token provider can be created with no parameters:

token_provider = hzkerberos.TokenProvider()

Authentication using a Keytab File

You can use a keytab file for retrieving the Kerberos ticket. In this case, the principal and full path of the keytab file must be specified:

token_provider = hzkerberos.TokenProvider(principal="jduke@EXAMPLE.COM", keytab="/etc/krb5.keytab")

Authentication using a Password

You can also use a password retrieving the Kerberos ticket. In this case, the principal and the password must be specified:

token_provider = hzkerberos.TokenProvider(principal="jduke@EXAMPLE.COM", password="s3cr3t")

Overriding the Generated Service Principal Name

A service principal name (SPN) has the following structure:

[SERVICE-NAME-PREFIX/][SERVICE-HOST][@REALM]

By default, the service principal name is generated automatically, using the following components:

• SERVICE-NAME-PREFIX: hz/
• SERVICE-HOST: Host or IP address of the member
• REALM: Blank

You can override SPN generation by specifying one or many of spn ,prefix and realm parameters:

token_provider = hzkerberos.TokenProvider(spn="my-service", prefix="hz", realm="EXAMPLE.COM")

Creating the Hazelcast Python Client

Once the token provider is created, you can pass it to the Hazelcast Python Client constructor. The token provider will be used by the client during authentication to the server.

client = hazelcast.HazelcastClient(token_provider=token_provider)

Server Configuration

Server security configuration (starting with 4.1) is documented in the Security section of the main Hazelcast documentation. Kerberos authentication is documented in the Security Reams sub-section.

The Kerberos support in Hazelcast has 2 configuration parts: identity and authentication. The identity part is responsible for retrieving the service ticket from Kerberos KDC (Key Distribution Center). The authentication part verifies the service tickets.

The following XML fragment can be used as an example of a working server configuration. However, it is recommended to read the completed documentation in order to fully understand the security aspects of Kerberos.

<security enabled="true">
    <member-authentication realm="kerberosRealm"/>
    <client-authentication realm="kerberosRealm"/>
    <realms>
        <realm name="kerberosRealm">
            <authentication>
                <kerberos>
                    <relax-flags-check>true</relax-flags-check>
                    <use-name-without-realm>false</use-name-without-realm>
                    <keytab-file>/common/krb5.keytab</keytab-file>
                    <principal>hz/${host}@EXAMPLE.COM</principal>
                </kerberos>
            </authentication>
            <identity>
                <kerberos>
                    <realm>EXAMPLE.COM</realm>
                    <keytab-file>/common/krb5.keytab</keytab-file>
                    <principal>hz/${host}@EXAMPLE.COM</principal>
                </kerberos>
            </identity>
        </realm>
    </realms>
</security>

Notes

1. Only the default cache is supported for storing/loading the Kerberos ticket. The default cache name is resolved using krb5_cc_default_name call.

Running the Tests

The Docker setup in this setup is strictly for test purposes.

Running the tests requires Docker Compose.

1. Put hazelcast-enterprise-4.2.2.jar, log4j-api-2.14.1.jar and log4j-core-2.14.1.jar in the docker folder. This is a temporary step to be removed in another iteration

2. Create an .env file with the Hazelcast Enterprise license key in the root of the project:

   HZ_LICENSEKEY=...
   

3. Run docker compose, which creates KDC/KAdmin and two Hazelcast Enterprise containers:

   docker-compose up
   

4. Build the Docker image for tests, this is required only once:

   docker build -t hazelcast-python-client-kerberos_app:latest -f app.Dockerfile .
   

5. Run the app container to run the tests whenever the code changes:

   docker run -it --rm --name hzkerberos_test \
       --env-file .env \
       --network=hazelcast-python-client-kerberos_hz \
       -v `pwd`:/home/hz/app \
       -v hazelcast-python-client-kerberos_common:/common \
       hazelcast-python-client-kerberos_app:latest test
   

Notes

1. When a container runs, it executes the corresponding default action, e.g., test for the app container. In order to get a shell instead of the default action, you can use the /bin/bash command.
2. If VERBOSE environment variable is 1, tracing is enabled for entrypoint scripts.
3. LOGGING_LEVEL environment variable for Hazelcast containers controls the logging level.

Accessing KDC

When the docker compose setup is running, you can access KDC by accessing its container:

docker-compose exec kdc /bin/bash

And starting kadmin.local:

rlwrap kadmin.local