openstack-manuals/doc/security-guide/ch027_storage.xml
Rodney Beede 973e7928c7 Wrote missing chapter on securing object storage (Swift).
Provides detailed guidelines on network and service security.

Change-Id: I7f6f13f710f27b56a444b038d06ad07a680cfc20
2013-12-18 17:41:33 +00:00

356 lines
18 KiB
XML
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<?xml version="1.0" encoding="UTF-8"?>
<chapter xmlns:xi="http://www.w3.org/2001/XInclude" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://docbook.org/ns/docbook" xmlns:db="http://docbook.org/ns/docbook" version="5.0" xml:id="ch027_storage"><?dbhtml stop-chunking?>
<title>Object Storage</title>
<para>OpenStack Object Storage (Swift) is a service that provides
storage and retrieval of data over HTTP. Objects (blobs of data)
are stored in an organizational hierarchy that offers anonymous
read-only access or ACL defined access based on the authentication
mechanism.</para>
<para>A consumer can store objects, modify them, or access them
using the HTTP protocol and REST APIs. Backend components of
Object Storage use different protocols for keeping the information
synchronized in a redundant cluster of services. For more details
on the API and the backend components see the
<link xlink:href="http://docs.openstack.org/api/openstack-object-storage/1.0/content/">OpenStack Storage documentation</link>.</para>
<para>For this document the components will be grouped into the
following primary groups:</para>
<para>
<orderedlist>
<listitem><para>Proxy services</para></listitem>
<listitem><para>Auth services</para></listitem>
<listitem><para>Storage services
<itemizedlist>
<listitem><para>Account service</para></listitem>
<listitem><para>Container service</para></listitem>
<listitem><para>Object service</para></listitem>
</itemizedlist>
</para></listitem>
</orderedlist>
</para>
<figure>
<title>An example diagram from the OpenStack Object Storage
Administration Guide (2013)</title>
<mediaobject>
<imageobject>
<imagedata contentdepth="329" contentwidth="494"
fileref="static/swift_network_diagram-1.png"
format="PNG" scalefit="1"/>
</imageobject>
</mediaobject>
</figure>
<note>
<para>
An Object Storage environment does not have to necessarily be
on the Internet and could also be a private cloud with the
"Public Switch" being part of the organizations internal
network infrastructure.
</para>
</note>
<section xml:id="ch027_storage-idpA">
<title>First thing to secure the network</title>
<para>The first aspect of a secure architecture design for
Object Storage is in the networking component. The Storage
service nodes use rsync between each other for copying data
to provide replication and high availability. In addition,
the proxy service communicates with the Storage service when
relaying data back and forth between the end-point client and
the cloud environment.</para>
<caution><para>None of these use any type of encryption or
authentication at this layer/tier.</para></caution>
<para>This is why you see a "Private Switch" or private
network ([V]LAN) in architecture diagrams. This data domain
should be separate from other OpenStack data networks as well.
For further discussion on security domains please see
<xref linkend="ch005_security-domains"/>.
</para>
<tip><para><emphasis>Rule:</emphasis> Use a private (V)LAN
network segment for your Storage services in the data domain.
</para></tip>
<para>This necessitates that the Proxy service nodes have dual
interfaces (physical or virtual):</para>
<orderedlist>
<listitem><para>One as a "public" interface for consumers
to reach</para></listitem>
<listitem><para>Another as a "private" interface with
access to the storage nodes</para></listitem>
</orderedlist>
<para>The following figure demonstrates one possible network
architecture.
</para>
<para>
<figure>
<title>Object storage network architecture with a
management node (OSAM)
</title>
<mediaobject>
<imageobject role="html">
<imagedata contentdepth="913"
contentwidth="1264"
fileref="static/swift_network_diagram-2.png"
format="PNG" scalefit="1"/>
</imageobject>
<imageobject role="fo">
<imagedata contentdepth="100%"
fileref="static/swift_network_diagram-2.png"
format="PNG" scalefit="1"
width="100%"/>
</imageobject>
</mediaobject>
</figure>
</para>
</section><!-- First thing to secure The Network -->
<section xml:id="ch027_storage-idpC1">
<title>Securing services general</title>
<section xml:id="ch027_storage-idpC12">
<title>Service runas user</title>
<para>It is recommended that you configure each service to
run under a non-root (UID 0) service account. One
recommendation is the username "swift" with primary group
"swift."</para>
</section>
<section xml:id="ch027_storage-idpC123">
<title>File permissions</title>
<para>/etc/swift contains information about the ring
topology and environment configuration. The following
permissions are recommended:</para>
<screen>
<prompt>#</prompt><userinput>chown -R root:swift /etc/swift/*</userinput>
<prompt>#</prompt><userinput>find /etc/swift/ -type f -exec chmod 640 {} \;</userinput>
<prompt>#</prompt><userinput>find /etc/swift/ -type d -exec chmod 750 {} \;</userinput>
</screen>
<para>This restricts only root to be able to modify
configuration files while allowing the services to read
them via their group membership in "swift."</para>
</section>
</section><!-- Securing Services General -->
<section xml:id="ch027_storage-idpD1">
<title>Securing storage services</title>
<para>The following are the default listening ports for the
various storage services:</para>
<informaltable>
<thead>
<tr>
<td>Service Name</td>
<td>Port</td>
<td>Type</td>
</tr>
</thead>
<tbody>
<tr>
<td>Account service</td>
<td>6002</td>
<td>TCP</td>
</tr>
<tr>
<td>Container service</td>
<td>6001</td>
<td>TCP</td>
</tr>
<tr>
<td>Object service</td>
<td>6000</td>
<td>TCP</td>
</tr>
<tr>
<td>Rsync</td>
<td>873</td>
<td>TCP</td>
</tr>
</tbody>
</informaltable>
<para>Authentication does not happen at this level in Object
Storage. If someone was able to connect to a Storage service
node on one of these ports they could access or modify data
without authentication. In order to secure against this issue
you should follow the recommendations given previously about
using a private storage network.</para>
<section xml:id="ch027_storage-idpD12">
<title>Object storage "account" terminology</title>
<para>An Object Storage "Account" is not a user account or
credential. The following explains the relations:</para>
<informaltable>
<tbody>
<tr>
<td>OpenStack Object Storage Account</td>
<td>Collection of containers; not user
accounts or authentication. Which users are
associated with the account and how they may
access it depends on the authentication system
used. See authentication systems later.
Referred to in this document as OSSAccount.
</td>
</tr>
<tr>
<td>OpenStack Object Storage Containers</td>
<td>Collection of objects. Metadata on the
container is available for ACLs. The meaning
of ACLs is dependent on the authentication
system used.</td>
</tr>
<tr>
<td>OpenStack Object Storage Objects</td>
<td>The actual data objects. ACLs at the
object level are also possible with metadata.
It is dependent on the authentication system
used.</td>
</tr>
</tbody>
</informaltable>
<tip>
<para>
<?dbhtml bgcolor="#DDFADE" ?><?dbfo bgcolor="#DDFADE" ?>
Another way of thinking about the above would be: A
single shelf (Account) holds zero or more -> buckets
(Containers) which each hold zero or more -> objects. A
garage (Object Storage cloud environment)
may have multiple shelves (Accounts) with each shelf
belonging to zero or more users.
</para>
</tip>
<para>At each level you may have ACLs that dictate who has
what type of access. ACLs are interpreted based on what
authentication system is in use. The two most common
types of authentication providers used are Keystone and
SWAuth. Custom authentication providers are also
possible. Please see the Object Storage Authentication
section for more information.</para>
</section>
</section><!-- Securing Storage Services -->
<section xml:id="ch027_storage-idpE1">
<title>Securing proxy services</title>
<para>A Proxy service node should have at least two interfaces
(physical or virtual): one public and one private. The
public interface may be protected via firewalls or service
binding. The public facing service is an HTTP web server that
processes end-point client requests, authenticates them, and
performs the appropriate action. The private interface does
not require any listening services but is instead used to
establish outgoing connections to storage service nodes on the
private storage network.</para>
<section xml:id="ch027_storage-idpE12">
<title>Use SSL/TLS</title>
<para>The built-in or included web server that comes with
Swift supports SSL, but it does not support transmission
of the entire SSL certificate chain. This causes issues
when you use a third party trusted and signed certificate
(ex: Verisign) for your cloud. The current work around is
to not use the built-in web server but an alternative web
server instead that supports sending both the public
server certificate as well as the CA signing authorities
intermediate certificate(s). This allows for end-point
clients that have the CA root certificate in their trust
store to be able to successfully validate your cloud
environments SSL certificate and chain. An example of
how to do this with mod_wsgi and Apache is given below.
Also consult the
<link xlink:href="http://docs.openstack.org/developer/swift/apache_deployment_guide.html">Apache Deployment Guide</link></para>
<screen>sudo apt-get install libapache2-mod-wsgi</screen>
<para>Modify file
<filename>/etc/apache2/envvars</filename> with</para>
<programlisting>
export APACHE_RUN_USER=swift
export APACHE_RUN_GROUP=swift
</programlisting>
<para>An alternative is to modify your Apache conf file
with</para>
<programlisting>
User swift
Group swift
</programlisting>
<para>Create a "swift" directory in your Apache document
root:</para>
<screen><prompt>#</prompt><userinput>sudo mkdir /var/www/swift/</userinput></screen>
<para>Create the file
<filename>$YOUR_APACHE_DOC_ROOT/swift/proxy-server.wsgi</filename>:</para>
<programlisting>from swift.common.wsgi import init_request_processor application, conf, logger, log_name = \
init_request_processor('/etc/swift/proxy-server.conf','proxy-server')</programlisting>
</section>
<section xml:id="ch027_storage-idpF1">
<title>HTTP listening port</title>
<para>You should run your Proxy service web server as a
non-root (no UID 0) user such as "swift" mentioned
before.
The use of a port greater than 1024 is required to make
this easy and avoid running any part of the web container
as root. Doing so is not a burden as end-point clients
are not typically going to type in the URL manually into
a web browser to browse around in the object storage.
Additionally, for clients using the HTTP REST API and
performing authentication they will normally
automatically grab the full REST API URL they are to use
as provided by the authentication response. OpenStacks
REST API allows for a client to authenticate to one URL
and then be told to use a completely different URL for
the actual service. Example: Client authenticates to
<uri>https://identity.cloud.example.org:55443/v1/auth</uri>
and gets a response with their authentication key and
Storage URL (the URL of the proxy nodes or load balancer)
of
<uri>https://swift.cloud.example.org:44443/v1/AUTH_8980</uri>.
</para>
<para>The method for configuring your web server to start
and run as a non-root user varies by web server and OS.
</para>
</section>
<section xml:id="ch027_storage-idpG1">
<title>Load balancer</title>
<para>If the option of using Apache is not feasible or for
performance you wish to offload your SSL work you may
employ a dedicated network device load balancer. This is
also the common way to provide redundancy and load
balancing when using multiple proxy nodes.</para>
<para>If you choose to offload your SSL ensure that the
network link between the load balancer and your proxy
nodes is on a private (V)LAN segment such that other nodes
on the network (possibly compromised) cannot wiretap
(sniff) the unencrypted traffic. If such a breach were to
occur the attacker could gain access to end-point client
or cloud administrator credentials and access the cloud
data.</para>
<para>The authentication service you use (e.g. Keystone,
SWAuth) will determine how you configure a different URL
in the responses to end-clients so they use your load
balancer instead of an individual Proxy service node.
</para>
</section>
</section><!-- Securing Proxy Services -->
<section xml:id="ch027_storage-idpH1">
<title>Object storage authentication</title>
<para>Object Storage uses wsgi to provide a middleware for
authentication of end-point clients. The authentication
provider defines what roles and user types exist. Some use
traditional username and password credentials while others may
leverage API key tokens or even client-side x.509 SSL
certificates. Custom providers can be integrated in using the
wsgi model.</para>
<section xml:id="ch027_storage-idpH12">
<title>Keystone</title>
<para>Keystone is the commonly used Identity provider in
OpenStack. It may also be used for authentication in
Object Storage. Coverage of securing Keystone is already
provided in <xref linkend="ch024_authentication"/>.</para>
</section>
<section xml:id="ch027_storage-idpH123">
<title>SWAuth</title>
<para>SWAuth is another alternative to Keystone.
In contrast to Keystone it stores the user accounts,
credentials, and metadata in object storage itself. More
information can be found on the SWAuth website at
<link xlink:href="http://gholt.github.io/swauth/">http://gholt.github.io/swauth/</link>.
</para>
</section>
</section><!-- Object Storage Authentication -->
<section xml:id="ch027_storage-idpI1">
<title>Other notable items</title>
<para>In /etc/swift/swift.conf on every service node there is
a "swift_hash_path_suffix" setting. This is provided to
reduce the chance of hash collisions for objects being stored
and avert one user overwriting the data of another user.
</para>
<para>This value should be initially set with a
cryptographically secure random number generator and
consistent across all service nodes. Ensure that it is
protected with proper ACLs and that you have a backup copy to
avoid data loss.</para>
</section>
</chapter>