hadoop-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Xine Jar <xineja...@googlemail.com>
Subject Re: problem with starting the Jobtracker and thenamenode
Date Wed, 08 Jul 2009 13:14:22 GMT
Thank you all for your great help,

I finally discovered a typing error in one of my configuration files.
Therefore my jobtracker was not able to bind the address to the port.

On the other hand I still have the Namenode which cannot be started. The log
file shows that the namenode is not formatted although I did this. I
appreciate it if someone can point to me the folders I can delete in order
to be able to reformat the node without deleting some critical information.
please note that I have deleted once the /tmp folder and tried to reformat
but this did not help !!

Vielen Dank,
CJ

On Wed, Jul 8, 2009 at 11:12 AM, Xine Jar <xinejar22@googlemail.com> wrote:

> Hallo,
> I am still struggling with the same problem. In order to be sure that IPv6
> is not creating a problem and that the JVM is not suffering from a bug, I
> have written a TCP client/server program in Java, I let the server run on
> the same machine that was creating the binding problem and using the same
> port number. Luckily or unfortunately, the Java server could bind the
> address normally to the port, the communication between the client and the
> server was successful, and the netstat command shows that the port is used.
>
> I order to be sure that the problem is not a bug for hadoop I have
> installed the version 0.20.0,
> copied the same configuration and still have the same problem.
>
> *I have few questions: *
> As I have previously mentioned, the namenode cannot be started and java
> complains that the node has not been formated. The other problem is with the
> jobtracker which is giving a binding error on the address:port.
>
> *Q1:* Is it possible that the second problem (jobtracker problem) is
> appearing because of the namenode problem?
>
> If it is a yes, how can I solve this? I have actually deleted the /tmp
> folder and reformated the node but the formatting error persists!! Do I have
> to do something else?
>
> If it is a no,  am I doing something stupid here in any of the
> configuration files?!!!
>
> *Q2:  *In order to kill all the instances of the dfs and the mapred, is it
> enough to execute the bin/stop-dfs.sh and bin/stop-mapred.sh on the
> namenode? Is it possible that the command line is showing that there is no
> jobtracker, no namenode, ...... but in fact the port is occupied?
>
> P.S: -Whether hadoop is running or not, the netstat command is showing
> always that the port
>          is free!!!!!
>        - My OS is a openSuse 9*
>
> *I am trying to debug everything, because I am somehow sick of it, It is
> certainly something stupid and most probably my fault (or my configuration
> or my way of running things) , since others managed to run hadoop on a
> cluster. I would appreciate any idea that helps to discover this the
> problem.
>
> Thank you,
> CJ
>
>
>
> On Tue, Jul 7, 2009 at 4:47 AM, Bogdan M. Maryniuk <
> bogdan.maryniuk@gmail.com> wrote:
>
>> On Mon, Jul 6, 2009 at 9:24 PM, Xine Jar<xinejar22@googlemail.com> wrote:
>> > 2.How can I check that ipv6 is really disabled on my JVM?
>>
>> You've mentioned it is Linux, but how do I know what distribution of
>> Linux's zoo you use?
>>
>> On Ubuntu it is sort of like this:
>> In the file "/etc/modprobe.d/aliases" find "alias net-pf-10 ipv6",
>> remove it and add the following:
>>
>> alias net-pf-10 off
>> alias ipv6 off
>>
>> Then reboot (welcome to the Linux).
>>
>> On RedHat and derivatives it is like this:
>> echo "alias net-pf-10 off" >>  /etc/modprobe.conf <ENTER>
>>
>> Then reboot (insert your ironic joke here). :-)
>>
>>
>> > and in which file can I insert the flag -Djava.net.preferIPv4Stack=true?
>> It is just a JVM parameter for a networking:
>> 1. http://java.sun.com/j2se/1.4.2/docs/guide/net/properties.html
>> 2.
>> http://java.sun.com/j2se/1.4.2/docs/guide/net/ipv6_guide/#ipv6-networking
>>
>> See $HADOOP/conf/hadoop-env.sh and look for HADOOP_OPTS that is
>> commented out by default.
>>
>> --
>> Kind regards, BM
>>
>> Things, that are stupid at the beginning, rarely ends up wisely.
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message