hadoop-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Nauroth <cnaur...@hortonworks.com>
Subject Re: Is it possible to turn on data node encryption without kerberos?
Date Thu, 07 Apr 2016 17:33:45 GMT
Hello Lin,

Great catch!  Yes, there is an implicit assumption that dfs.block.access.token.enable would
have to be set to true.

The main purpose of this property is to activate usage of block access tokens for client DataNode
interactions.  For example, upon opening a file, the user authenticates to the NameNode, the
NameNode issues block access tokens that declare the user is authorized to access the blocks
of the file, and then the client presents those block access tokens to the DataNodes to authorize
access to the block.

The implicit dependency for data transfer protocol encryption comes about because that feature
also relies on the same infrastructure within the NameNode for management of encryption keys.
 Without the property enabled, the NameNode doesn't activate its secret management infrastructure,
and therefore there is no encryption key available to use for data transfer protocol encryption.

Most fully secured deployments would set dfs.block.access.token.enable=true as part of the
full security enablement procedure.  Since you're not fully enabling security, your case is
the first time that I've seen this happen.

--Chris Nauroth

From: Lin Zhao <lin@exabeam.com<mailto:lin@exabeam.com>>
Date: Wednesday, April 6, 2016 at 7:48 PM
To: Chris Nauroth <cnauroth@hortonworks.com<mailto:cnauroth@hortonworks.com>>,
"musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>" <musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>>,
"user@hadoop.apache.org<mailto:user@hadoop.apache.org>" <user@hadoop.apache.org<mailto:user@hadoop.apache.org>>
Subject: Re: Is it possible to turn on data node encryption without kerberos?


I followed the TestEncryptedTransfer test case and set dfs.block.access.token.enable to true
and this issue is resolved. I can't find documentation this property is mandatory for encrypted
data transfer. What does this property do?

From: Chris Nauroth <cnauroth@hortonworks.com<mailto:cnauroth@hortonworks.com>>
Date: Wednesday, April 6, 2016 at 4:02 PM
To: "musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>" <musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>>,
Lin Zhao <lin@exabeam.com<mailto:lin@exabeam.com>>, "user@hadoop.apache.org<mailto:user@hadoop.apache.org>"
Subject: Re: Is it possible to turn on data node encryption without kerberos?

It is possible to turn on data transfer protocol encryption without enabling Kerberos authentication.
 We have a test suite in the Hadoop codebase named TestEncryptedTransfer that configures data
transfer encryption, but not Kerberos, and those tests are passing.

The hadoop.rpc.protection setting is unrelated to data transfer protocol.  Instead, it controls
the SASL quality of protection for the RPC connections used by many Hadoop client/server interactions.
 This won't really be active unless Kerberos authentication is enabled though.

Please note that even though it's possible to enable data transfer protocol encryption without
using Kerberos authentication in the cluster, the benefit of that is questionable in a production
deployment.  Without Kerberos authentication, it's very easy for an unauthenticated user to
spoof another user and access their HDFS files.  Whether or not the data is encrypted in transit
becomes irrelevant at that point.

--Chris Nauroth

From: Musty Rehmani <musty_rehmani@yahoo.com.INVALID<mailto:musty_rehmani@yahoo.com.INVALID>>
Reply-To: "musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>" <musty_rehmani@yahoo.com<mailto:musty_rehmani@yahoo.com>>
Date: Wednesday, April 6, 2016 at 2:54 PM
To: Lin Zhao <lin@exabeam.com<mailto:lin@exabeam.com>>, "user@hadoop.apache.org<mailto:user@hadoop.apache.org>"
Subject: Re: Is it possible to turn on data node encryption without kerberos?

Kerberos is used to authenticate user or service principal to grant access to cluster. It
doesn't encrypt data blocks coming in and out of cluster.
Sent from Yahoo Mail on Android<https://overview.mail.yahoo.com/mobile/?.src=Android>

On Wed, Apr 6, 2016 at 4:36 PM, Lin Zhao
<lin@exabeam.com<mailto:lin@exabeam.com>> wrote:
I've been trying to secure block data transferred by HDFS. I added below to hdfs-site.xml
and core-site xml to the data node and name node and restart both.



When I try to put a file from the hdfs command line shell, the operation fails with "connection
is reset" and I see following from the datanode log:

"org.apache.hadoop.hdfs.server.datanode.DataNode: Failed to read expected encryption handshake
from client at / Perhaps the client is running an older version of Hadoop
which does not support encryption"

I am able to reproduce this on two different deployments. I was following https://hadoop.apache.org/docs/r2.7.2/hadoop-project-dist/hadoop-common/SecureMode.html#Authentication,
but didn't turn on kerberos authentication. No authentication works in my environment. Can
this be the reason the handshake fails?

Any help is appreciated.


Lin Zhao

View raw message