Thursday, October 18, 2012

Secure your cluster with kerberos

Many times I get questions about a safe and fast way to secure a cluster without big steps like integrate AD structures, simply to prevent unauthorized access. I created this writeup to let you know the steps you need. I used CentOS 6.3 and CDH 4.0.1, but you can use other distributions as well.

Setup KDC on a Linux Box

Install kerberos5 related packages as well as kadmin, too. First thing you have to do is to replace EXAMPLE.COM, which is delivered per default, with your own realm. I used ALO.ALT here.

Example config:

# hadoop1> cat /etc/krb5.conf 
 default_realm = ALO.ALT
 dns_lookup_realm = false
 dns_lookup_kdc = false
  ALO.ALT = {
  kdc = HADOOP1.ALO.ALT:88
  admin_server = HADOOP1.ALO.ALT:749
  default_domain = HADOOP1.ALO.ALT
 .alo.alt = ALO.ALT
 alo.alt = ALO.ALT

  kdc = FILE:/var/log/krb5kdc.log
  admin_server = FILE:/var/log/kadmin.log
  default = FILE:/var/log/krb5lib.log

Now tweak your DNS or /etc/hosts to reflect the settings, if you use /etc/hosts be sure you've deployed this both files across your nodes (hosts as well as krb5.conf).

# hadoop1> cat /etc/hosts hadoop1.alo.alt             hadoop1 hadoop2.alo.alt    hadoop2   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6

Enable ACL's by writing rules into kdc's config space:
# hadoop1> cat /var/kerberos/krb5kdc/kadm5.acl
*/admin@ALO.ALT *

Now create your kerberos random key with "kdb5_util create -s" and start kerberos as well as kadmin. If you face some errors, revisit your config.

Adding Principals

To use kerberos, you have to add some principals, since we're talking about hadoop I setup all needed princs. We use kadmin.local on the server which hosts the KDC.

addprinc -randkey hdfs/hadoop1.alo.alt@ALO.ALT
addprinc -randkey mapred/hadoop1.alo.alt@ALO.ALT
addprinc -randkey yarn/hadoop1.alo.alt@ALO.ALT
addprinc -randkey hbase/hadoop1.alo.alt@ALO.ALT

addprinc -randkey HTTP/hadoop1.alo.alt@ALO.ALT

You have to create a password for your <YOUR_USERNAME>, that's all. Now you should be able to sudo into your user account and kinit yourself (su - <YOUR_USERNAME> && kinit).

You need to export the keytabs for all services you want to secure (in fact, all services you want to start in your cluster). We need keytabs for HDFS, MapReduce, YARN and HBase in this case. We start with HDFS (always from kadmin's CLI)

xst -norandkey -k hdfs.keytab hdfs/hadoop1.alo.alt@ALO.ALT HTTP/hadoop1.alo.alt@ALO.ALT
xst -norandkey -k mapred.keytab mapred/hadoop1.alo.alt@ALO.ALT HTTP/hadoop1.alo.alt@ALO.ALT
xst -norandkey -k yarn.keytab yarn/hadoop1.alo.alt@ALO.ALT HTTP/hadoop1.alo.alt@ALO.ALT

Now you have all keytabs exported into the directory where you are:

# hadoop1> ls
hdfs.keytab mapred.keytab yarn.keytab

Chmod the files into the correct owner (chmod 400 hdfs:hadoop hdfs.keytab, chmod 400 mapred:hadoop mapred.keytab and so on and copy them in the right place into $HADOOP_HOME/conf/). From now on, all steps are pretty easy and well documented, so I post only the necessary config changes.

Enable Kerberos in a Cluster

I writeup the changes only, if you haven't them in your *.xml, wrap the XML notation around (<property><name>NAME<value>VALUE</value></name></property>).


dfs.block.access.token.enable = true
dfs.namenode.keytab.file = <PATH/TO/hdfs.keytab>
dfs.namenode.kerberos.principal = hdfs/_HOST@ALO.ALT
dfs.namenode.kerberos.internal.spnego.principal = HTTP/_HOST@ALO.ALT
dfs.secondary.namenode.keytab.file = <PATH/TO/hdfs.keytab>
dfs.secondary.namenode.kerberos.principal = hdfs/_HOST@ALO.ALT
dfs.secondary.namenode.kerberos.internal.spnego.principal = HTTP/_HOST@ALO.ALT = 700
dfs.datanode.address =
dfs.datanode.http.address =
dfs.datanode.keytab.file = <PATH/TO/hdfs.keytab>
dfs.datanode.kerberos.principal = hdfs/_HOST@ALO.ALT
dfs.webhdfs.enabled = true
dfs.web.authentication.kerberos.principal = HTTP/_HOST@ALO.ALT
dfs.web.authentication.kerberos.keytab = <PATH/TO/hdfs.keytab>

Startup the Namenode, watch the logs for issues and try out if you can connect:
"hadoop dfs -ls /"

It's important that we set a sticky bit to /tmp:
# hadoop1> sudo -u hdfs kinit -k -t hdfs.keytab hdfs/hadoop1.alo.alt@ALO.ALT
# hadoop1> sudo -u hdfs hadoop fs -chmod 1777 /tmp


mapreduce.jobtracker.kerberos.principal = mapred/_HOST@ALO.ALT
mapreduce.jobtracker.keytab.file = mapred/_HOST@ALO.ALT
mapreduce.tasktracker.kerberos.principal = mapred/_HOST@ALO.ALT
mapreduce.tasktracker.keytab.file = mapred/_HOST@ALO.ALT
mapred.task.tracker.task-controller = org.apache.hadoop.mapred.LinuxTaskController = mapred

Create a file called taskcontroller.cfg in $HADOOP_HOME/conf/

# hadoop1> cat /etc/hadoop/conf/taskcontroller.cfg 

Now startup the Tasktracker as well as the Jobtracker, too. Watch the logs for issues, and if all is going well up start a sample MR job (like pi or something else) as a "normal" user for testing purposes. Please note, you've to kinit first. To control that you have a gotten a valid ticket use klist.