Internal security

This document covers network-level security. For documentation on the authentication and authorization features introduced in Riak 2.0, see Authentication and Authorization and Managing Security Sources

This article discusses standard configurations and port settings to use when providing network security for a Riak Cluster. There are two classes of access control for Riak:

• Other Riak nodes participating in the cluster
• Clients making use of the Riak cluster

The settings for both access groups are located in your cluster’s configuration settings. If you are using the newer configuration system, you can set a host and port for each node in that node’s riak.conf file, setting listener.protobuf if you are using Riak’s Protocol Buffers interface or listener.http if you are using HTTP (or listener.https if you are using SSL). If you are using the older configuration system, adjust the settings of pb, http, or https, depending on which client interface you are using.

Make note of these configurations and set up your firewall to allow incoming TCP access to those ports or IP address/port combinations. Exceptions to this are the handoff_ip and handoff_port directives. Those are for communication between Riak nodes only.

Inter-node Communication

Riak uses the Erlang distribution mechanism for most inter-node communication. Riak identifies other machines in the ring using Erlang identifiers (<hostname or IP>, e.g. riak@10.9.8.7). Erlang resolves these node identifiers to a TCP port on a given machine via the Erlang Port Mapper daemon (epmd) running on each cluster node.

By default, epmd binds to TCP port 4369 and listens on the wildcard interface. For inter-node communication, Erlang uses an unpredictable port by default; it binds to port 0, which means the first available port.

For ease of firewall configuration, Riak can be configured to instruct the Erlang interpreter to use a limited range of ports. For example, to restrict the range of ports that Erlang will use for inter-Erlang node communication to 6000-7999, add the following lines to the configuration file on each Riak node:

erlang.distribution.port_range.minimum = 6000
erlang.distribution.port_range.maximum = 7999

{ kernel, [
            {inet_dist_listen_min, 6000},
            {inet_dist_listen_max, 7999}
          ]},

The above lines should be added into the top level list in app.config, at the same level as all the other applications (e.g. riak_core). Then configure your firewall to allow incoming access to TCP ports 6000 through 7999 from whichever network(s) contain your Riak nodes.

Riak Node Ports

Riak nodes in a cluster need to be able to communicate freely with one another on the following ports:

• epmd listener: TCP:4369
• handoff_port listener: TCP:8099
• range of ports specified in app.config or riak.conf

Riak Client Ports

Riak clients must be able to contact at least one machine in a Riak cluster on the following TCP ports:

Riak Search Ports

Riak’s search feature relies on Apache Solr, which runs on each Riak node if security has been enabled. When Riak’s Search subsystem starts up, JMX opens a well-known port as well as some ephemeral ports. The well-known port is determined by the value of the search.solr.jmx_port in each node’s Search configuration. The default is 8985.

In addition to JMX ports, Solr also binds to a well-known port of its own, as determined by each node’s search.solr.port setting, which is also located in each node’s Search configuration. The default is 8093.

Riak Security Community

Riak

Riak is a powerful open-source distributed database focused on scaling predictably and easily, while remaining highly available in the face of server crashes, network partitions or other (inevitable) disasters.

Commitment

Data security is an important and sensitive issue to many of our users. A real-world approach to security allows us to balance appropriate levels of security and related overhead while creating a fast, scalable, and operationally straightforward database.

Continuous Improvement

Though we make every effort to thwart security vulnerabilities whenever possible (including through independent reviews), no system is completely secure. We will never claim that Riak is 100% secure (and you should seriously doubt anyone who claims their solution is). What we can promise is that we openly accept all vulnerabilities from the community. When appropriate, we’ll publish and make every attempt to quickly address these concerns.

Balance

More layers of security increase operational and administrative costs. Sometimes those costs are warranted, sometimes they are not. Our approach is to strike an appropriate balance between effort, cost, and security.

For example, Riak does not have fine-grained role-base security. Though it can be an attractive bullet-point in a database comparison chart, you’re usually better off finely controlling data access through your application or a service layer.

Notifying Riak

If you discover a potential security issue, please email us at security@riak.com, and allow us 48 hours to reply.

We prefer to be contacted first, rather than searching for blog posts over the Internet. This allows us to open a dialogue with the security community on how best to handle a possible exploit without putting any users at risk.

Security Best Practices

Authentication and Authorization

For instructions on how to apply permissions and to require client authentication, please see our documentation on Riak Security.

Network Configurations

Being a distributed database means that much of Riak’s security springs from how you configure your network. We have a few recommendations for Security and Firewalls.

Client Auth

All of the Riak client libraries support encrypted TCP communication as well as authentication and authorization. For instructions on how to apply permissions and to require client authentication, please see our documentation on Riak Security.

Multi-Datacenter Replication

For those versions of Riak that support Multi Data Center (MDC) Replication, you can configure Riak 1.2+ to communicate over SSL, to seamlessly encrypt the message traffic.

See also: Multi Data Center Replication: SSL in the Enterprise Documentation.