Hmmmm ... no luck. I've tried now with whirr.bootstrap-user,
whirr.login-user (which whirr tells me is now deprecated in favor of
whirr.bootstrap-user), and with no user at all. Result is always the
same: instances start up, but whirr fails to initialize them. The
whirr log shows the following error message (which apparently indicates
an ssh login failure):
java.lang.IllegalStateException: Invalid packet: indicated length
1349281121 too large
And on the nodes, I keep seeing Ubuntu starting up a process that echo's
the following message: Please login as the user "ubuntu" rather than
the user "root".
So it seems that somehow whirr keeps trying to login as root and
failing. Not sure why.
The Ubuntu AMI I'm using is based off of one provided by the Ubuntu
project itself. Specifically ami-4fad7426. (See this page for more
details: http://cloud.ubuntu.com/ami/)
Anyone have any idea what might be going wrong here?
Thanks,
DR
On 09/16/2012 03:48 PM, Sebastian Schoenherr wrote:
Hi David,
try to set: whirr.login-user=ubuntu and skip the bootstrap-user.
works for me,
Chers Sebsatian
On 16/09/2012 16:41, David Rosenstrauch wrote:
Whoops sorry - didn't understand. Props file posted below Again,
this is using Whirr v0.8.
DR
---
whirr.env.repo=cdh4
whirr.cluster-name=hadoopcc
whirr.instance-templates=1 hadoop-namenode+hadoop-jobtracker,8
hadoop-datanode+hadoop-tasktracker
whirr.instance-templates-minimum-number-of-instances=1
hadoop-namenode+hadoop-jobtracker, 6 hadoop-datanode+hadoop-tasktracker
whirr.max-startup-retries=4
whirr.provider=aws-ec2
whirr.identity=...
whirr.credential=...
whirr.private-key-file=${sys:user.home}/.ssh/id_rsa
whirr.public-key-file=${sys:user.home}/.ssh/id_rsa.pub
whirr.hadoop.install-function=install_cdh_hadoop
whirr.hadoop.configure-function=configure_cdh_hadoop
whirr.hardware-id=cc1.4xlarge
whirr.location-id=us-east-1
whirr.image-id=us-east-1/ami-dd7fcfb4
whirr.bootstrap-user=ubuntu
hadoop-hdfs.dfs.replication=2
hadoop-mapreduce.mapred.tasktracker.map.tasks.maximum=8
hadoop-mapreduce.mapred.tasktracker.reduce.tasks.maximum=8
hadoop-mapreduce.mapred.reduce.tasks=64
hadoop-mapreduce.mapred.task.timeout=1800000
hadoop-mapreduce.mapred.child.java.opts=-Xmx1024m
On 09/15/2012 04:20 PM, Alex Heneveld wrote:
Hi David,
I think Andrei is asking if you can send the props/config file
(recipe) you
are using.
Best,
Alex
On Sep 15, 2012 7:43 PM, "David Rosenstrauch" <[email protected]> wrote:
We're using whirr v0.8.
What message is it that you're suggesting I add to the
hadoop.properties
file?
Thanks,
DR
On Sat, September 15, 2012 12:59 pm, Andrei Savu wrote:
Please add to this message your .properties file. What version of
Whirr
are
you using?
On Sep 15, 2012 7:52 PM, "David Rosenstrauch" <[email protected]>
wrote:
We've been using whirr to launch instances of the (redhat-based)
Amazon
AMI, which has been working. However, I'm now trying to switch to
launching an ubuntu-based AMI, and I'm running into problems.
Ubuntu, as you might recall, doesn't allow you to log in directly as
root.
Rather, you log in as user ubuntu, and then sudo to root. This is
causing problems when whirr tries to set up my data nodes.
I've configured whirr to use "ubuntu" as the bootstrap user, which
seems
to work fine initially. But after a short while in the installation
process, it fails. And when I go onto the data node machine to watch
what's happening, I see that something (most likely whirr) is
trying to
login as root. So I'm fairly certain that's what's causing the
installation to fail.
Is there a way around this issue?
Thanks,
DR
