Support Questions
Find answers, ask questions, and share your expertise

Problem starting nodemanager

Contributor

Hello all

 

I could not start YARN service on my fresh installation.

 

I followed following URL for the installation and I am using cm-5.4.3

 

http://www.cloudera.com/content/cloudera/en/documentation/core/latest/topics/cm_ig_install_path_c.ht...

Installation Path C - Manual Installation Using Cloudera Manager Tarballs

 

I am using one node for startup.

 

I got following errors while startup YARN first time.

 

 

1.jpg

 

 

 

Details screen

 

2.jpg

more details

 

3.jpg

 

stderr output

 

 

4.jpg

 

 

stdout output

 

5.jpg

 

 

Role log details

 

6.jpg

 

 

 

my /var/log/directory has link to /opt/cloudera/log/cloudera-scm-server directory as follows

 

7.jpg

 

I could create a file in cloudera-scm-server's log directory as cloudera-scm user (I changed the user's login capabilities for testing)

 

8.jpg

 

A python process listening 9000/TCP

 

9.jpg

 

/opt/cloudera/lib and /opt/cloudera/log directories have following directories 

 

 

10.jpg

 

I could also have read/write option on hadoop-yarn directory as yarn user, I tested it

 

I changed /tmp with 

mount -o remount exec /tmp 

No progress, then I revert it to original state

 

So how I install and start YARN and Nodemanager

 

PS:Our Unix team has some standarts about FS sizing, therefore I create my own log and lib directories and I linked into /var/lib and /var/log as subdir.

 

Best Regards

Murat

 

 

 

13 REPLIES 13

Contributor

Hello all

 

Has anyone any idea about this issue ?

 

Thanks

Murat

Cloudera Employee

We would need the /var/run/cloudera-scm-agent/process/ directory stderr and stdout logs for the process that's failing for us to understand the issue. 

 

You're using the CM tarball, which could have the /var/run/cloudera-scm-agent directory configured if you have specific limitations. 

You can define the location here: tarball_root/etc/init.d/cloudera-scm-agent

 

Do you have passwordless sudo access for the single user? 

 

The docs are located here: http://www.cloudera.com/content/cloudera/en/documentation/core/latest/topics/install_singleuser_reqt...

Contributor

Hello

 

Thank you very much for your reply

 

If I understand clearly you are mentioning about defining following mount point in scm-agent configuration

 

cm_processes 132189892 65384 132124508 1% /opt/cloudera/cloudera-manager/cm-5.4.3/run/cloudera-scm-agent/process

 

[root@<hostname>:/opt/cloudera/cloudera-manager/cm-5.4.3/etc/init.d] ll
total 20
-rwxr-xr-x 1 cloudera-scm cloudera-scm 6656 Jun 26 07:30 cloudera-scm-agent
-rwxr-xr-x 1 cloudera-scm cloudera-scm 7328 Jun 26 07:30 cloudera-scm-server
-rwxr-xr-x 1 cloudera-scm cloudera-scm 2894 Jun 26 07:30 cloudera-scm-server-db

 

How can I define the process directory in cloudera-scm-agent file?

 

cloudera-scm user could not have sudo access, we are using Centrify's dzdo and it has different mech.

 

Thanks

msuluhan

Cloudera Employee

The user either needs passwordless sudo, or a whitelist of sudo commands. I'm not too familiar with Centrify. 

You will need sudo access to enable security, which is why it is a requirement to either have passwordless sudo for the user or a whitelist for the user. 

 

I don't see the stderr / stdout file to determine why the installation was failing.