HDFS Forum

Namenode not starting (6x EC2 micro instance)

  • #42422
    Manuel
    Participant

    Hello!

    I installed a 6 node cluster on EC2 micro instances, based on this guide:
    http://hortonworks.com/kb/ambari-on-ec2/

    Yes, the guide recommends to use medium and large (for NN) instances.
    As this is just a proof of concept install I thought to try it out on micro instances.
    If this is the simple reason for the problem I am happy to go with a larger instance.

    Everything worked fine so far until service startup, where the namenode is constantly failing.
    A link to the namenode log:
    http://pastebin.com/xzj8275k

    Any feedback would be very much appreciated!
    Thanks!

    Best,

    Manuel

to create new topics or reply. | New User Registration

  • Author
    Replies
  • #42435
    Robert Molina
    Moderator

    Hi Manuel,
    What version of HDP did you install? What is the OS you installed on EC2 to host hadoop? Can you start the service manually by executing the following:

    su – hdfs -c ‘/usr/lib/hadoop/bin/hadoop-daemons.sh start namenode’

    Regards,
    Robert

    #42440
    Chris Nauroth
    Moderator

    Hi Manuel,

    I expect that an EC2 micro instance won’t work, at least for the NameNode. Micro instances only have 0.6 GB of RAM. The NameNode generally needs a lot more RAM than that. I believe Ambari configures the NameNode heap size to 1 GB by default, which might explain why the process fails to start.

    Hope this helps,
    –Chris

    #42441
    Manuel
    Participant

    Hi Robert, hi Chris,

    many thanks for the fast reply!

    When starting the namenode via the console:

    [root@ambarimaster bin]# su -l hdfs -c “/usr/lib/hadoop/bin/hadoop-daemon.sh –config /etc/hadoop/conf start namenode”
    starting namenode, logging to /var/log/hadoop/hdfs/hadoop-hdfs-namenode-ambarimaster.hdp.hadoop.out
    Java HotSpot(TM) 64-Bit Server VM warning: MaxNewSize (655360k) is equal to or greater than the entire heap (262144k). A new generation size of 262080k will be used.
    Error occurred during initialization of VM
    Could not reserve enough space for object heap

    This sound to me like the hint from Chris, doesn’t it?
    As this is purely a test installation; where can I limit the MaxNewSize?
    Or is the “Error occurred during initialization of VM” a different problem apart from that?

    Thanks!

    Manuel

    #42443
    Manuel
    Participant

    Hi Robert,

    sorry, the info you asked for:
    AMI ID: RHEL-6.4_GA-x86_64-10-Hourly2 (ami-75342c01)
    HDFS 1.2.0.1.3.2.0
    MapReduce 1.2.0.1.3.2.0
    Nagios 03.05.2000
    Ganglia 03.05.2000
    Hive 0.11.0.1.3.2.0
    Pig 0.11.1.1.3.2.0
    Sqoop 1.4.3.1.3.2.0

    Best,

    Manuel

    #43451
    Robert Molina
    Moderator

    Hi Manuel,
    I believe Chris is on it. Please try to provision a larger machine where Namenode can use around 1GB of memory.

    Regards,
    Robert

The topic ‘Namenode not starting (6x EC2 micro instance)’ is closed to new replies.

Support from the Experts

A HDP Support Subscription connects you experts with deep experience running Apache Hadoop in production, at-scale on the most demanding workloads.

Enterprise Support »

Become HDP Certified

Real world training designed by the core architects of Hadoop. Scenario-based training courses are available in-classroom or online from anywhere in the world

Training »

Hortonworks Data Platform
The Hortonworks Data Platform is a 100% open source distribution of Apache Hadoop that is truly enterprise grade having been built, tested and hardened with enterprise rigor.
Get started with Sandbox
Hortonworks Sandbox is a self-contained virtual machine with Apache Hadoop pre-configured alongside a set of hands-on, step-by-step Hadoop tutorials.
Modern Data Architecture
Tackle the challenges of big data. Hadoop integrates with existing EDW, RDBMS and MPP systems to deliver lower cost, higher capacity infrastructure.