HA Nameservice using Cloudera Director

[#1]

I'm trying to use Cloudera Director to create our cluster. We have 2 masters and I want to have NameNode on both of them with HA enabled.

I can't find anything in the documentation about HA and specifing the name of the nameservice in the documentation and the git script at https://github.com/cloudera/director-scripts/blob/master/high-availability/aws.enableha.sample.conf doesn't really look like it makes sense.

Is there any documentation on writing the conf file to support HA?

Thanks,

Owen

[#2 · Accepted solution]

Hi Owen,

Thanks for trying out Director.

Have you read the README.md co-located with the conf file you linked?
https://github.com/cloudera/director-scripts/tree/master/high-availability/README.md
This git repo provides scripts (written in python or groovy) that you can
use to enable HA for HDFS on a newly created cluster. The scripts,
https://github.com/cloudera/director-scripts/blob/master/high-availability/python/enable-hdfs-ha.py
or
https://github.com/cloudera/director-scripts/blob/master/high-availability/groovy/enableHdfsHa.groov...
allow you to specify the nameservice on the command line.

Director 1.5 and Cloudera Manager 5.4 do not support pre-configuring HDFS
for HA at cluster creation time. The provided scripts will invoke a command
on Cloudera Manager to enable HA for an existing HDFS service. Director
will not have any knowledge about the role changes that result from these
scripts, but could still be used for growing and shrinking worker nodes.
Full HA support at cluster creation time is being worked on for a future
release of Director.

David

[#3]

Hi David

Thats embarrassing - I went straight to the script and didn't register there was more to it.

Thanks for coming back to me, great help.

Owen