Just remove the directories configured for dfs.name.dir and dfs.data.dir and run the `hadoop namenode -format` again.

On 3/27/13 11:31 AM, Aji Janis wrote:
well... I found this in the datanode log

ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: java.io.IOException: Incompatible namespaceIDs in /opt/hadoop-data/hadoop/hdfs/data: namenode namespaceID = 2089335599; datanode namespaceID = 1868050007




On Wed, Mar 27, 2013 at 11:23 AM, Eric Newton <[email protected] <mailto:[email protected]>> wrote:

    "0 live nodes"  that will continue to be a problem.

    Check the datanode logs.

    -Eric


    On Wed, Mar 27, 2013 at 11:20 AM, Aji Janis <[email protected]
    <mailto:[email protected]>> wrote:


        I removed everything
        under /opt/hadoop-data/hadoop/hdfs/data/current/ because it
        seemed like old files were hanging around and I had to remove
        them before I can start re-initialization.


        I didn't move anything to /tmp or try reboot.
        my old accumulo instance had everything under /accumulo (in
        hdfs) and its still there but i m guessing me deleting stuff
        from hadoop-data has deleted a bunch of its stuff.

        i tried to restart zookeeper and hadoop and it came up fine
        but now my namenode url says there 0 live nodes (instead of 5
        in my cluster). Doing a ps -ef | grep hadoop on each node in
        cluster however shows that hadoop is running.... so i am not
        sure what I messed up. Suggestions?

        Have I lost accumulo for good? Should I just recreate the
        instance?


        On Wed, Mar 27, 2013 at 10:52 AM, Eric Newton
        <[email protected] <mailto:[email protected]>> wrote:

            Your DataNode has not started and reported blocks to the
            NameNode.

            Did you store things (zookeeper, hadoop) in /tmp and
            reboot?  It's a common thing to do, and it commonly
            deletes everything in /tmp.  If that's the case, you will
            need to shutdown hdfs and run:

            $ hadoop namenode -format

            And then start hdfs again.

            -Eric


            On Wed, Mar 27, 2013 at 10:47 AM, Aji Janis
            <[email protected] <mailto:[email protected]>> wrote:

                I see thank you. When I bring up hdfs (start-all from
                node with jobtracker) I see the following message on
                url: http://mynode:50070/dfshealth.jsp

                *Safe mode is ON. /The ratio of reported blocks 0.0000
                has not reached the threshold 0.9990. Safe mode will
                be turned off automatically./
                **2352 files and directories, 2179 blocks = 4531
                total. Heap Size is 54 MB / 888.94 MB (6%) *
                *
                *
                Whats going on here?



                On Wed, Mar 27, 2013 at 10:44 AM, Eric Newton
                <[email protected] <mailto:[email protected]>>
                wrote:

                    This will (eventually) delete everything created
                    by accumulo in hfds:

                    $ hadoop fs -rmr /accumulo

                    Accumulo will create a new area to hold your
                    configurations.  Accumulo will basically abandon
                    that old configuration.  There's a class that can
                    be used to clean up old accumulo instances in
                    zookeeper:

                    $ ./bin/accumulo
                    org.apache.accumulo.server.util.CleanZookeeper
                    hostname:port

                    Where "hostname:port" is the name of one of your
                    zookeeper hosts.

                    -Eric



                    On Wed, Mar 27, 2013 at 10:29 AM, Aji Janis
                    <[email protected] <mailto:[email protected]>> wrote:

                        Thanks Eric. But shouldn't I be cleaning up
                        something in the hadoop-data directory too?
                        and Zookeeper?



                        On Wed, Mar 27, 2013 at 10:27 AM, Eric Newton
                        <[email protected]
                        <mailto:[email protected]>> wrote:

                            To re-initialize accumulo, bring up
                            zookeeper and hdfs.

                            $ hadoop fs -rmr /accumulo
                            $ ./bin/accumulo init

                            I do this about 100 times a day on my dev
                            box. :-)

                            -Eric


                            On Wed, Mar 27, 2013 at 10:10 AM, Aji
                            Janis <[email protected]
                            <mailto:[email protected]>> wrote:

                                Hello,

                                We have the following set up:

                                zookeeper - 3.3.3-1073969
                                hadoop - 0.20.203.0
                                accumulo - 1.4.2

                                Our zookeeper crashed for some reason.
                                I tried to doing a clean stop of
                                everything and then brought up (in
                                order) zookeeper and hadoop (cluster).
                                But when trying to do a start-all on
                                accumulo I get the following message
                                gets infinitely printed to the screen:

                                “26 12:45:43,551 [server.Accumulo]
                                INFO : Waiting for accumulo to be
                                initialized”



                                Doing some digging on the web it seems
                                that accumulo is hosed and needs some
                                re-intialization. It also appears that
                                may be I need to clean out things from
                                zookeeper and hadoop prior to a
                                re-initialization. Has any one done
                                this before? Can someone please
                                provide me some directions on what to
                                do (or not to do)? Really appreciate
                                help on this. Thanks.










Reply via email to