pig-user  

Re: upgrade problem (16.4 -> 17.1)

Gert Pfeifer
Wed, 02 Jul 2008 21:13:18 -0700

It is running now. The thing is, that the data nodes seem to convert
their data into a new format/version or something...
This took a very long time. It would have been nice to see some hint in
the log of the data nodes, like "please wait while upgrading".

Gert

Olga Natkovich schrieb:
Try and restart your cluster. If this does not help, post a message to
[EMAIL PROTECTED] Olga
-----Original Message-----
From: Gert Pfeifer [EMAIL PROTECTED] Sent: Wednesday, July 02, 2008 1:50 AM
To: pig-user@incubator.apache.org
Subject: upgrade problem (16.4 -> 17.1)

I did the upgrade now, but there might be a problem... or it just takes very long time.

The name node is running in safe mode, but there are no data nodes.

I tried this:
$ ./hadoop dfsadmin -upgradeProgress status

Upgrade for version -13 has been completed.
Upgrade is not finalized.

The datanodes say:
----------------------
STARTUP_MSG: Starting DataNode
STARTUP_MSG:   host = dns-file-1/10.3.61.211
STARTUP_MSG:   args = []
STARTUP_MSG:   version = 0.17.1
STARTUP_MSG:   build =
http://svn.apache.org/repos/asf/hadoop/core/branches/branch-0.
17 -r 669344; compiled by 'hadoopqa' on Thu Jun 19 01:18:25 UTC 2008 ************************************************************/ 2008-07-02 10:34:56,154 INFO org.apache.hadoop.dfs.Storage: Recovering storage directory /var/data/hadoop/dfs/data from previous upgrade. 2008-07-02 10:35:07,105 INFO org.apache.hadoop.dfs.Storage: Upgrading storage directory /var/tmp/hadoop/dfs/data.
   old LV = -11; old CTime = 0.
   new LV = -13; new CTime = 1214984938285
----------------------

The namenode loops in the log with:
----------------------
The ratio of reported blocks 0.0000 has not reached the threshold 0.9990. Safe mode will be turned off automatically. org.apache.hadoop.dfs.SafeModeException: Cannot renew lease for DFSClient_-18804611. Name node is in safe mode.
----------------------

Why is there no datanode connecting? Did I forget something? I did not try to finalize the upgrade yet. I guess that wouldn't be a good idea.

My problem might be similar to this:
http://mail-archives.apache.org/mod_mbox/hadoop-core-user/2007
09.mbox/<[EMAIL PROTECTED]>
 - but it doesn't give me a solution

Thanks for any hints,
Gert

Gert Pfeifer wrote:
I'm using 16.4.

Gert

Olga Natkovich wrote:
The latest code works with Hadoop 17. What version do you have?

Olga

-----Original Message-----
From: Gert Pfeifer [EMAIL PROTECTED]
Sent: Tuesday, July 01, 2008 11:01 AM
To: pig-user@incubator.apache.org
Subject: Re: connecting to cluster with bin/pig

Hi,
exectype is not touched, so it defaults to mapreduce
(correct me if
I am
wrong)

My mistake was to put the hadoop-site.xml into the path. Thank you very much!!!

Now I found this:
Protocol org.apache.hadoop.dfs.ClientProtocol version mismatch. (client = 29, server = 23)

Is there a compatibility mode, or do I have to update the server?

Gert

Alan Gates schrieb:
Are you setting exectype to 'mapreduce'? It looks like
you have it
set to 'local'.  Also, make sure you include the directory
that your
hadoop-site.xml is in in the class path, not the file itself.

If neither of those resolve your issue, please post your whole pig.properties file.

Alan.

Gert Pfeifer wrote:
Hi,
I am trying to start pig for the first time, so here is a
beginner's
question.

How do I tell the bin/pig shell script where the cluster
can be found?
I used the conf/pic.properties as follows:

# clustername, name of the hadoop jobtracker. If no port
is defined
port 50020 will be used.
cluster=<ip address of the job tracker>

Then I get a message:
[main] INFO

org.apache.pig.backend.hadoop.executionengine.HExecutionEngine -
Connecting to hadoop file system at: file:///

Then I get the grunt shell on the local file system,
which is not
quite what I wanted.

I also tried this:
java -cp pig.jar:../../path/to/hadoop-site.xml org.apache.pig.Main

But I saw the same result. So how do I connect to the name
node and
the job tracker? I guess I need both, don't I?

Thanks for any hints,
Gert