Reply
Expert Contributor
Posts: 192
Registered: ‎07-01-2015

Cloudera Director first run cluster fail

[ Edited ]

Hi,

 running Cloudera Director 2.8 I am not able to spin up even a simple non-ha 1-3-1 cluster without manual fixing of service settings. Issues like permission denied for JobHistory (tries to create /user/history in HDFS) or Hive (tries to access /tmp in HdFS) links to the not-prepared default HDFS directories and their permissions.

 

My question is: is this a bug related to the CDH 5.15 / Director 2.8 or should I take care of these directories mysefl?

And if yes, where is the best point to inject the

"hdfs -mkdir /user/" 

"hdfs -chown hdfs:hadoop /user"

"hdfs -chmod 775... "

...

 

commands in the director template?

 

Is there a possiblity to execute something in-between the Services during the first run (i.e. NameNode come up, but YARN(with Jobhistory) not yet)?

 

Thanks

Cloudera Employee
Posts: 60
Registered: ‎02-18-2014

Re: Cloudera Director first run cluster fail

Hi Tomas79,

 

Which cloud provider are you using? If you are using a configuration file, does it work in previous versions of Cloudera Director?

 

We don't have hooks at the granularity that you asked about, but they shouldn't be necessary. Those kinds of typical setup actions are handled by Cloudera Manager.

Expert Contributor
Posts: 192
Registered: ‎07-01-2015

Re: Cloudera Director first run cluster fail

Provider: AWS

Did not tried previous version of the Cloudera Director. Will try to run again..

 

Announcements