AnsweredAssumed Answered

Starting mapr-warden on second node causes CLDB  to fail on first node

Question asked by greesil on Mar 8, 2014
Latest reply on Mar 11, 2014 by greesil
I have two nodes:

Node1 (10.10.11.33): cldb  fileserver  hbinternal  hbregionserver  nfs  tasktracker  zookeeper<br>
Node2 (10.10.11.36): fileserver  hbinternal  hbregionserver  tasktracker

 * After I start up warden on node 1, it works fine.  I can install license, run a MapReduce job, etc, but it's all just on node 1.
 * As soon as I run "sudo service mapr-warden start" on node 2, node 1 stops working.
 * Error message: **ERROR (10009) -  Couldn't connect to the CLDB service**


Any ideas why this might be happening?

I run this one the first node:
<code>
--------
ZKSERVERS="10.10.11.33"

CLDBSERVERS="10.10.11.33"

CLUSTERNAME="cluster1"

service mapr-zookeeper stop

service mapr-warden stop

apt-get --force-yes -y --purge remove  mapr-zookeeper mapr-zk-internal

apt-get --force-yes -y install mapr-zookeeper mapr-zk-internal

rm -rf /opt/mapr/zkdata/*

echo "" > /opt/mapr/conf/disktab

/opt/mapr/server/configure.sh -C $ZKSERVERS -Z $CLDBSERVERS -N $CLUSTERNAME

/opt/mapr/server/disksetup -F ~/drives.txt

service mapr-zookeeper start

sleep 15
service mapr-zookeeper qstatus
</code>
------


On the second node:
---------
<code>
ZKSERVERS="10.10.11.33"

CLDBSERVERS="10.10.11.33"

CLUSTERNAME="cluster1"


service mapr-warden stop

echo "" > /opt/mapr/conf/disktab

/opt/mapr/server/configure.sh -C $ZKSERVERS -Z $CLDBSERVERS -N $CLUSTERNAME

/opt/mapr/server/disksetup -F ~/drives.txt

-----------

$: /opt/mapr/bin/maprcli node cldbmaster
cldbmaster                                                 
         
ServerID: 5742312317975256079 HostName: hadoop1.blahblah.com

</code>

Outcomes