FAQ
(adding back scm-users)

On Tue, Jul 16, 2013 at 1:16 PM, Darren Lo wrote:

Hi Tim,

If you don't have anything running on the cluster yet, then it's probably
easiest to just stop your cluster, delete everything in the data
directories, and delete your cluster, then go through the add cluster
wizard again to set things up how you want.

If you want to move roles around instead, then for each role that supports
decommissioning (such as datanode and regionserver), you can select the
Decommission action one by one. Navigate to the page for that role instance
and select the Decommission command from the Actions menu. If you don't see
the action, then the role can't / doesn't need to be Decommissioned. You
may need to read other documentation for the right way to move roles like
this, such as the NameNode. It's potentially fairly complex, which is why I
recommend setting up your cluster from scratch with the right host
assignments if you can.

If you want to decommission the entire host (which will not affect
Cloudera Manager), you can select the host from the Hosts page and select
the action to decommission everything on the host.

Gateways don't take any resources. They just help keep track of which
hosts should get client configuration when you run the Deploy Client
Configuration command.

Thanks,
Darren

On Tue, Jul 16, 2013 at 12:56 PM, Tim Washburn wrote:

Hi Darren,

Thanks for the reply. If I elect to decomm the manager from being a
datanode, regionserver, gateway etc will still operate as my manager? I ask
as the particular machine was selected as the manager based on disk size
and memory. Thinking that the manager's requirements weren't that intensive
when compared to the Namenodes and datanodes. Yes it is small and a POC to
wet the appetite of the business - it may grow up to become a real cluster
some day :--)

Regards
Tim Washburn
Global IT Security
Bio-Rad Laboratories, Inc
(510)741-6888



From: Darren Lo <dlo@cloudera.com>
To: Tim Washburn <tim_washburn@bio-rad.com>
Date: 07/16/2013 11:35 AM
Subject: Re: Cluster Setup with Cloudera Manger on with samll
cluster - determining what services/roles for wach member of the cluster.
------------------------------



Hi Tim,

During initial installation you can pick which hosts have which roles by
clicking Inspect Role Assignments. CM will automatically suggest
assignments, which is what happened on your cluster.

If you want to move roles around, you can use the decommission command to
accomplish this, but there's probably nothing wrong with your current
cluster setup. If you find that you don't have enough RAM on your namenode
host to run your roles well, then you can consider moving some roles off of
that host. Note that I'm assuming this is not a production cluster since
there are so few nodes.

Thanks,
Darren


On Mon, Jul 15, 2013 at 3:49 PM, Tim Washburn <*tim_washburn@bio-rad.com*<tim_washburn@bio-rad.com>>
wrote:

Hi All,

I've just finished setting up a small cluster consisting of the following:

1 single server for the Cloudera manager
2 servers for the primary and secondary namenodes
3 servers for the datanodes.

I used the default templates and allowed CM to determine what each node
would be. For the most part it looks like CM chose the correct items per
member. My question is how to tell if the primary namenode has all the
correct services running? This question also applies to the other nodes as
well. Part of my concern is that the Manager is also a datanode is this
correct? Can I have the Clodera manage be just a manager and not a datanode?

+++++++++++++++++++++++++++++++++++++++++++++
The primary namenode has the the following services running:

master (usherlxsnn1)
regionserver (usherlxsnn1)
datanode (usherlxsnn1)
namenode (usherlxsnn1)
gateway (usherlxsnn1)
hivemetastore (usherlxsnn1)
beeswax_server (usherlxsnn1)
hue_server (usherlxsnn1)
impalad (usherlxsnn1)
statestore (usherlxsnn1)
jobtracker (usherlxsnn1)
tasktracker (usherlxsnn1)
activitymonitor (usherlxsnn1)
alertpublisher (usherlxsnn1)
eventserver (usherlxsnn1)
hostmonitor (usherlxsnn1)
servicemonitor (usherlxsnn1)
oozie_server (usherlxsnn1)
sqoop_server (usherlxsnn1)
server (usherlxsnn1)

The secondary namenode has these services:

regionserver (usherlxsnn2)
datanode (usherlxsnn2)
secondarynamenode (usherlxsnn2)
gateway (usherlxsnn2)
impalad (usherlxsnn2)
tasktracker (usherlxsnn2)

The all the datanodes have the following services:

regionserver (usherlxsdn1)
datanode (usherlxsdn1)
gateway (usherlxsdn1)
impalad (usherlxsdn1)
tasktracker (usherlxsdn1)

The manager has the following services:

regionserver (usherlxsmgr1)
datanode (usherlxsmgr1)
gateway (usherlxsmgr1)
impalad (usherlxsmgr1)
tasktracker (usherlxsmgr1)
+++++++++++++++++++++++++++++++++

Regards
Tim Washburn



--
Thanks,
Darren


--
Thanks,
Darren


--
Thanks,
Darren

Search Discussions

Discussion Posts

Previous

Related Discussions

Discussion Navigation
viewthread | post
posts ‹ prev | 2 of 2 | next ›
Discussion Overview
groupscm-users @
categorieshadoop
postedJul 15, '13 at 10:49p
activeJul 16, '13 at 8:18p
posts2
users2
websitecloudera.com
irc#hadoop

2 users in discussion

Darren Lo: 1 post Tim Washburn: 1 post

People

Translate

site design / logo © 2022 Grokbase