CDH version is none

[#1]

Hi.

I am trying to install 3 datanodes and one manager to a pseudo cluster. Everything appears to work OK but I get the 3 datanodes say CDH Version is none (the manger chub is OK).
The errors in my log are:

>>[12/Jun/2014 14:59:56 +0000] 10950 MainThread agent ERROR Failed to connect to previous supervisor.
...

>>error: [Errno 111] Connection refused

Could it be something to do with my etc/hosts:
127.0.0.1    localhost
192.168.56.1    chub
192.168.56.2    vm1
192.168.56.3    vm2
192.168.56.4    vm3

same on all 4 machines?

I've been stuck on this for a week or so, so any help is MUCH APPRECIATED!!!

more info:

host -v -t A `hostname`

on the vms gives me not found error but hostname gives correct answer e.g. vm1

I'm using the user 'vm' which is on each machine with NOPASSWD entry in /etc/sudoers and chub can passwordless ssh to each vm@vm1, vm@vm2, vm@vm3.

also CM warns for each vm:

The Cloudera Manager Agent's process directory is on a filesystem with less than 200.0 MiB of its space free. /run/cloudera-scm-agent (free: 199.4 MiB (99.62%), capacity: 200.1 MiB)

details: all ubuntu 64 bit 12.04, CDH 5

[#2]

Hi there.

First, on each host, make sure this returns your FQDN and IP:

python -c "import socket; print socket.getfqdn(); print socket.gethostbyname(socket.getfqdn())"

If that does not return the fqdn and ip, then, in your hosts file, make sure you have the FQDN before host for each IP.

<IP>  <FQDN> <host>

That said, the errors you mention do not necessarily mean there is something wrong.

Your concern seems to be that "CDH Version" is none.  Where are you seeing that, exactly?  Can you manage the hosts via Cloudera Manager?

Ben

[#3]

Thanks for your reply. I'm not sure of the correct terminology. I can see the hosts but cannot assign roles to them. Their fqdn's are just the hostname. I think CDH is not installing on them but I'm not sure what to look for on them to confirm this.

[#4]

The python script gives me e.g. vm3 192.168.56.4, which is what I would expect.

[#5 · Accepted solution]

I uninstalled the cloudera installer then installed it in only vm1-3, which installed cdh5 onto them. Then after a few times of deleting the cluster and re-adding hosts all 4 now work.