You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-user@hadoop.apache.org by Ilya Vishnevsky <Il...@e-legion.com> on 2007/06/07 14:32:54 UTC
Why datanode does not work properly on slave?
Hello! I'm deploying Nutch on two computers. When I run start-all.sh
script all goes good but data node on slave computer does not log
anything. All other parts of Hadoop (namenode, jobtracker, both
tasktrackers and datanode on master) log their information properly.
Also, when I put some files from local file system into hadoop fs they
are put only into master's data folder. Slave's data folder is empty.
At the same time when I run stop-all.sh script I get message, that
slave's datanode is being stopped. It means that it has been running
before.
Do you know what may cause this problem?
Re: Why datanode does not work properly on slave?
Posted by Konstantin Shvachko <sh...@yahoo-inc.com>.
> Failed to create file ... on client 127.0.0.1 because this cluster has no datanodes.
Check the ip addresses.
If the name-node machine has 2 network cards the name-node may start on
one of them but
the data-nodes will try to contact the other.
There is something wrong with the scripts on Windows. We should file a bug.
I checked that direct calls of
bin/hadoop-daemon.sh start namenode
instead of start-dfs actually work fine.
Don't know about your logging problem. Did you check "log4j.properties"?
--Konstantin
Ilya Vishnevsky wrote:
>Well, I've updated version of hadoop to 0.13. The exception is not
>thrown anymore but the problem still exists. Datanode on the slave is
>shown to be running, but it does not log anything nor keep any data
>within itself. If "slaves" file contains both localhost (master) and
>slave's IP, then "hadoop fs -put ..." command puts information only in
>the master's datanode. If "slaves" file contains only slave's IP then I
>get message like following:
> Failed to create file ... on client 127.0.0.1 because this cluster has
>no datanodes. It seems as if I missed to configurate something in hadoop
>but I don't know what.
> Help please!
>
>
>-----Original Message-----
>From: Dhruba Borthakur [mailto:dhruba@yahoo-inc.com]
>Sent: Thursday, June 07, 2007 8:46 PM
>To: hadoop-user@lucene.apache.org
>Subject: RE: Why datanode does not work properly on slave?
>
>This Arithmetic Exception that you are seeing might be related to the
>bug
>http://issues.apache.org/jira/browse/HADOOP-1122. This bug has been
>fixed in
>release 0.13.
>
>Thanks,
>Dhruba
>
>-----Original Message-----
>From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
>Sent: Thursday, June 07, 2007 5:52 AM
>To: hadoop-user@lucene.apache.org
>Subject: RE: Why datanode does not work properly on slave?
>
>I've just changed dfs.replication property in the hadoop-site.xml from 1
>to 2. Now I get "ArithmeticException: / by" zero when try to put
>something into dfs.
> Also I want to say that both nodes are on Windows.
>
>
>
>-----Original Message-----
>From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
>Sent: Thursday, June 07, 2007 4:33 PM
>To: nutch-user@lucene.apache.org
>Subject: Why datanode does not work properly on slave?
>
> Hello! I'm deploying Nutch on two computers. When I run start-all.sh
>script all goes good but data node on slave computer does not log
>anything. All other parts of Hadoop (namenode, jobtracker, both
>tasktrackers and datanode on master) log their information properly.
> Also, when I put some files from local file system into hadoop fs they
>are put only into master's data folder. Slave's data folder is empty.
> At the same time when I run stop-all.sh script I get message, that
>slave's datanode is being stopped. It means that it has been running
>before.
> Do you know what may cause this problem?
>
>
>
>
RE: Why datanode does not work properly on slave?
Posted by Ilya Vishnevsky <Il...@e-legion.com>.
Well, I've updated version of hadoop to 0.13. The exception is not
thrown anymore but the problem still exists. Datanode on the slave is
shown to be running, but it does not log anything nor keep any data
within itself. If "slaves" file contains both localhost (master) and
slave's IP, then "hadoop fs -put ..." command puts information only in
the master's datanode. If "slaves" file contains only slave's IP then I
get message like following:
Failed to create file ... on client 127.0.0.1 because this cluster has
no datanodes. It seems as if I missed to configurate something in hadoop
but I don't know what.
Help please!
-----Original Message-----
From: Dhruba Borthakur [mailto:dhruba@yahoo-inc.com]
Sent: Thursday, June 07, 2007 8:46 PM
To: hadoop-user@lucene.apache.org
Subject: RE: Why datanode does not work properly on slave?
This Arithmetic Exception that you are seeing might be related to the
bug
http://issues.apache.org/jira/browse/HADOOP-1122. This bug has been
fixed in
release 0.13.
Thanks,
Dhruba
-----Original Message-----
From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
Sent: Thursday, June 07, 2007 5:52 AM
To: hadoop-user@lucene.apache.org
Subject: RE: Why datanode does not work properly on slave?
I've just changed dfs.replication property in the hadoop-site.xml from 1
to 2. Now I get "ArithmeticException: / by" zero when try to put
something into dfs.
Also I want to say that both nodes are on Windows.
-----Original Message-----
From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
Sent: Thursday, June 07, 2007 4:33 PM
To: nutch-user@lucene.apache.org
Subject: Why datanode does not work properly on slave?
Hello! I'm deploying Nutch on two computers. When I run start-all.sh
script all goes good but data node on slave computer does not log
anything. All other parts of Hadoop (namenode, jobtracker, both
tasktrackers and datanode on master) log their information properly.
Also, when I put some files from local file system into hadoop fs they
are put only into master's data folder. Slave's data folder is empty.
At the same time when I run stop-all.sh script I get message, that
slave's datanode is being stopped. It means that it has been running
before.
Do you know what may cause this problem?
RE: Why datanode does not work properly on slave?
Posted by Dhruba Borthakur <dh...@yahoo-inc.com>.
This Arithmetic Exception that you are seeing might be related to the bug
http://issues.apache.org/jira/browse/HADOOP-1122. This bug has been fixed in
release 0.13.
Thanks,
Dhruba
-----Original Message-----
From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
Sent: Thursday, June 07, 2007 5:52 AM
To: hadoop-user@lucene.apache.org
Subject: RE: Why datanode does not work properly on slave?
I've just changed dfs.replication property in the hadoop-site.xml from 1
to 2. Now I get "ArithmeticException: / by" zero when try to put
something into dfs.
Also I want to say that both nodes are on Windows.
-----Original Message-----
From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
Sent: Thursday, June 07, 2007 4:33 PM
To: nutch-user@lucene.apache.org
Subject: Why datanode does not work properly on slave?
Hello! I'm deploying Nutch on two computers. When I run start-all.sh
script all goes good but data node on slave computer does not log
anything. All other parts of Hadoop (namenode, jobtracker, both
tasktrackers and datanode on master) log their information properly.
Also, when I put some files from local file system into hadoop fs they
are put only into master's data folder. Slave's data folder is empty.
At the same time when I run stop-all.sh script I get message, that
slave's datanode is being stopped. It means that it has been running
before.
Do you know what may cause this problem?
RE: Why datanode does not work properly on slave?
Posted by Ilya Vishnevsky <Il...@e-legion.com>.
I've just changed dfs.replication property in the hadoop-site.xml from 1
to 2. Now I get "ArithmeticException: / by" zero when try to put
something into dfs.
Also I want to say that both nodes are on Windows.
-----Original Message-----
From: Ilya Vishnevsky [mailto:Ilya.Vishnevsky@e-legion.com]
Sent: Thursday, June 07, 2007 4:33 PM
To: nutch-user@lucene.apache.org
Subject: Why datanode does not work properly on slave?
Hello! I'm deploying Nutch on two computers. When I run start-all.sh
script all goes good but data node on slave computer does not log
anything. All other parts of Hadoop (namenode, jobtracker, both
tasktrackers and datanode on master) log their information properly.
Also, when I put some files from local file system into hadoop fs they
are put only into master's data folder. Slave's data folder is empty.
At the same time when I run stop-all.sh script I get message, that
slave's datanode is being stopped. It means that it has been running
before.
Do you know what may cause this problem?