HDFS Forum

Issue: Non dfs space

  • #30336
    Member

    I have a 3 node cluster with 1 namenode and 2 datanodes.
    For some reason on one of the datanode the non dfs is filling up.
    When I check the Name node UI I see
    Total(gb) Used(gb) Non-dfs used Remaining(gb) Remaining(%) Blocks
    datanode 1 2 In Service 125.05 17.15 77.91 29.98 23.98 7956
    datanode2 0 In Service 125.05 25.99 98.86 0.19 0.15 7881

    When I do a df on the datanode2
    Filesystem Size Used Avail Use% Mounted on
    /dev/mapper/VolGroup00-LogVol00 43G 41G 66M 100% /

    Can you please help out.
    – Thanks.

to create new topics or reply. | New User Registration

  • Author
    Replies
  • #30350
    Sasha J
    Moderator

    Check logs directory, it may be filled up.

    Thank you!
    Sasha

    #30355
    Member

    I checked the logs and it seems to be okay
    [mapred@qa-centos-amd64-28 /]$ du -h /var/log/hadoop/
    24K /var/log/hadoop/mapred/userlogs/job_201303061716_0597
    28K /var/log/hadoop/mapred/userlogs
    6.8M /var/log/hadoop/mapred
    14M /var/log/hadoop/hdfs
    21M /var/log/hadoop/

    The difference I notice on two datanodes

    datanode1 had
    [mapred@qa-centos-amd64-27 hdfs]$ du -sh /grid/hadoop/hdfs/dn/current/
    5.9G /grid/hadoop/hdfs/dn/current/
    [mapred@qa-centos-amd64-27 hdfs]$ du -sh /grid1/hadoop/hdfs/dn/current/
    5.7G /grid1/hadoop/hdfs/dn/current/
    [mapred@qa-centos-amd64-27 hdfs]$ du -sh /grid2/hadoop/hdfs/dn/current/
    5.7G /grid2/hadoop/hdfs/dn/current/

    And on datanode2
    [mapred@qa-centos-amd64-28 /]$ du -sh /grid/hadoop/hdfs/dn/current/
    9.1G /grid/hadoop/hdfs/dn/current/
    [mapred@qa-centos-amd64-28 /]$ du -sh /grid1/hadoop/hdfs/dn/current/
    8.6G /grid1/hadoop/hdfs/dn/current/
    [mapred@qa-centos-amd64-28 /]$ du -sh /grid2/hadoop/hdfs/dn/current/
    8.4G /grid2/hadoop/hdfs/dn/current/

    Doesn’t these have to be same. What should I do to balance this? how do I clear off stuff from the second datanode.

    #30410
    Robert
    Participant

    Hi Member,
    I don’t think balancing will help, since I believe the files are not related to what is being stored in the cluster, but most likely temp and log files. You can try the following to try and find which files are occupying the most space.

    run at the root level du -cs * | sort -g which give an output of folders and their file sizes. Go into the folder that has a large size and run the command again until you get into a folder where you can isolate which file or set of files are occupying all the space.

    Hope that helps.
    Regard,
    Robert

The topic ‘Issue: Non dfs space’ 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.