Apache
/
hbase
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HBASE-27226 Document native TLS support in Netty RPC (#4717) 

Signed-off-by: Balazs Meszaros <meszibalu@apache.org>
Signed-off-by: Bryan Beaudreault <bbeaudreault@apache.org>
This commit is contained in:
Andor Molnár 2022-09-08 19:12:45 +02:00 committed by GitHub
parent 2bfa540be5
commit 793f020e13
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 309 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

309
src/main/asciidoc/_chapters/security.adoc
View File

@ -675,6 +675,315 @@ For more information about ACLs, please see the <<hbase.accesscontrol.configurat
It should be possible for clients to authenticate with the HBase cluster through the REST gateway in a pass-through manner via SPNEGO HTTP authentication.
This is future work.
== Transport Level Security (TLS) in HBase RPC communication
Since version `2.6.0` HBase supports TLS encryption in server-client and Master-RegionServer communication.
link:https://en.wikipedia.org/wiki/Transport_Layer_Security/[Transport Layer Security (TLS)] is a standard
cryptographic protocol designed to provide communications security over a computer network. HBase TLS implementation
works exactly how secure websites are accessed via *https* prefix in a web browser: once established all communication
on the channel will be securely hidden from malicious access.
The encryption works at the transport level which means it's independent of the configured authentication method. Secure
client access mentioned in the previous section requires Kerberos to be configured and used in HBase authentication, while
TLS can be configured with any other SASL mechanism or even with simple client access methods, effectively preventing
attackers from eavesdropping the communication. No Kerberos KDC or other complicated infrastructure required.
HBase TLS is based on the Netty library therefore it only works with Netty client and server RPC implementations. Netty's
powerful SSL implementation is a great foundation for highly secure and performant communication providing the latest and
greatest cryptographic solution at all times.
Since Region Servers effectively work as clients from Master's perspective, TLS supports encrypted communication
between cluster members too.
[NOTE]
====
From version 2.6.0 HBase supports the
link:https://hadoop.apache.org/docs/r3.3.4/hadoop-project-dist/hadoop-common/CredentialProviderAPI.html[Hadoop CredentialProvider API]
to avoid storing sensitive information in HBase configuration files. The recommended way of storing keystore / truststore passwords
is to use one of the supported credential providers e.g. the local jceks file provider. You can find more information
about how to setup credential providers in the Hadoop documentation.
The CLI interface for accessing the Hadoop Credential Shell is also available in HBase CLI. Type `hbase credential` to get help.
====
=== Server side configuration
We need to set up Java key store for the server. Key store is the list of private keys that a server can use to configure TLS
encryption. See link:https://en.wikipedia.org/wiki/Transport_Layer_Security/[TLS wikipedia page]
for further details of the protocol. Add the following configuration to `hbase-site.xml` on Master, Region Servers and HBase
clients:
[source,xml]
----
<property>
<name>hbase.server.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.rpc.tls.keystore.location</name>
<value>/path/to/keystore.jks</value>
</property>
----
Use `hbase.rpc.tls.keystore.password` alias to retrieve key store password from Hadoop credential provider.
NOTE: The supported storefile formats are based on the registered security providers and the loader can be autodetected from
the file extension. If needed, the file format can be explicitly specified with the `hbase.rpc.tls.keystore.type` property.
=== Client side configuration
We need to configure trust store for the client. Trust store contains the list of certificates that the client should trust
when doing the handshake with the server. Add the following to `hbase-site.xml`.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.rpc.tls.truststore.location</name>
<value>/path/to/truststore.jks</value>
</property>
----
Use `hbase.rpc.tls.truststore.password` alias to retrieve trust store password from Hadoop credential provider.
NOTE: The supported storefile formats are based on the registered security providers and the loader can be autodetected from
the file extension. If needed, the file format can be explicitly specified with the `hbase.rpc.tls.truststore.type` property.
However, specifying a trust store is not always required. Standard JDK implementations are shipped with a standard list
of trusted certificates (the certificates of Certificate Authorities) and if your private key is provided by one of them,
you don't need to configure your clients to trust it. Similarly to an internet browser, you don't need to set up the
certificates of every single website you're planning to visit. Later in this documentation we'll walk through the steps of
creating self-signed certificates which requires a trust store setup.
You can check the list of public certificate authorities shipped with your JDK implementation:
----
keytool -keystore $JAVA_HOME/jre/lib/security/cacerts -list
----
Password is empty by default.
=== Creating self-signed certificates
While obtaining globally trusted certificates from Certificate Authorities is convenient, it's perfectly valid to generate
your own private/public keypairs and set them up specifically for the HBase cluster. Especially if you don't want to enable
public access to the cluster, paying money for a certificate doesn't make sense.
Follow the following steps to generate self-signed certificates.
. Create SSL key store JKS to store local credentials
Please note that the alias (-alias) and the distinguished name (-dname) must match the hostname of the machine that is
associated with, otherwise hostname verification won't work.
----
keytool -genkeypair -alias $(hostname -f) -keyalg RSA -keysize 2048 -dname "cn=$(hostname -f)" -keypass password -keystore keystore.jks -storepass password
----
At the end of this operation you'll have as many key store files as many servers you have in your cluster. Each cluster member
will have its own key store.
[start=2]
. Extract the signed public key (certificate) from each key store
----
keytool -exportcert -alias $(hostname -f) -keystore keystore.jks -file $(hostname -f).cer -rfc
----
[start=3]
. Create SSL trust store JKS containing certificates for the clients
The same truststore (storing all accepted certs) should be shared on participants of the cluster. You need to use different
aliases to store multiple certificates in the same truststore. Name of the aliases doesn't matter.
----
keytool -importcert -alias [host1..3] -file [host1..3].cer -keystore truststore.jks -storepass password
----
=== Upgrading existing non-TLS cluster with no downtime
Here are the steps needed to upgrade an already running HBase cluster to TLS without downtime by taking advantage of
port unification functionality. There's a property on server side called `hbase.server.netty.tls.supportplaintext`
which makes possible to accept TLS and plaintext connections on the same socket port.
. Create the necessary key stores and trust stores for all server participants as described in the previous section.
. Enable secure communication on the Master node in _server-only_ mode with plaintext support.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabled</name>
<value>false</value>
</property>
<property>
<name>hbase.server.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.server.netty.tls.supportplaintext</name>
<value>true</value>
</property>
...keystore / truststore setup ...
----
Restart the Master. Master now accepts both TLS/non-TLS connections and works with non-TLS in client mode.
[start=3]
. Enable secure communication on the Region Servers in both _server and client_ mode with plaintext support.
Client mode here will ensure that RegionServer's communication to Master is encrypted.
[WARNING]
====
*Replication*
If you have read replicas enabled in your cluster or replication between two different clusters, you
have to break this into two steps. Secure communication has to be enabled on the _server side_ first with plaintext
support and once all Region Servers are upgraded you can repeat the upgrade by enabling _client side_ as well.
You have to prepare all Region Servers for secure communication before upgrading the client side.
====
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.server.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.server.netty.tls.supportplaintext</name>
<value>true</value>
</property>
...keystore / truststore setup ...
----
Restart Region Servers in rolling restart fashion. They send requests with TLS and accept both TLS and non-TLS communication.
[start=4]
. Enable secure communication on the clients.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabled</name>
<value>true</value>
</property>
...truststore setup ...
----
[start=5]
. Enable client-mode TLS on master and disable plaintext mode.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.server.netty.tls.enabled</name>
<value>true</value>
</property>
<property>
<name>hbase.server.netty.tls.supportplaintext</name>
<value>false</value>
</property>
----
Restart Master.
[start=6]
. Disable plaintext communication on the Region Servers by removing `supportplaintext` property. Restart RSs in rolling
restart fashion.
WARNING: Once `hbase.client.netty.tls.enabled` is enabled on the server side, the cluster will only be able to communicate
with other clusters which have TLS enabled. For example, this would impact inter-cluster replication.
=== Additional configuration
==== Enabled protocols
Comma-separated list of TLS protocol versions to enable. Default is empty.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.enabledProtocols</name>
<value>TLSv1.2,TLSv1.3</value>
</property>
----
==== Default protocol
Set the default TLS protocol version to use. Default is TLSv1.2. Use this protocol if enabled protocols is not defined.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.protocol</name>
<value>TLSv1.2</value>
</property>
----
==== Enabled cipher suites
List of enabled cipher suites in TLS protocol. Useful when you want to disable certain cipher suites due to recent
security concerns. Default value is a mix of CBC and GCM ciphers. Due to performance reasons we prefer CBC ciphers for
Java 8 and GCM ciphers for Java 9+.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.ciphersuites</name>
<value>TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256</value>
</property>
----
==== Certificate Revocation Checking
There's a built-in mechanism in JDK's TrustManager which automatically checks certificates for revocation. See
link:https://docs.oracle.com/cd/E19263-01/817-5215/ssl.html#wp19807[Managing Server Certificates]. Disabled by default.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.clr</name>
<value>false</value>
</property>
----
==== Online Certificate Status Protocol
Enables link:https://en.wikipedia.org/wiki/OCSP_stapling[OCSP] stapling. Please note that not all `SSLProvider`
implementations support OCSP stapling and an exception will be thrown upon. Disabled by default.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.ocsp</name>
<value>false</value>
</property>
----
==== Client handshake timeout
Set the TLS client handshake timeout is milliseconds. Default is 5 seconds.
[source,xml]
----
<property>
<name>hbase.client.netty.tls.handshaketimeout</name>
<value>5000</value>
</property>
----
== Securing Access to HDFS and ZooKeeper
Secure HBase requires secure ZooKeeper and HDFS so that users cannot access and/or modify the metadata and data from under HBase. HBase uses HDFS (or configured file system) to keep its data files as well as write ahead logs (WALs) and other data. HBase uses ZooKeeper to store some metadata for operations (master address, table locks, recovery state, etc).