NodeManager fails to start

[#1]

Hello folks!

I have a cluster with 9 machines, running on CDH 6.2 (OnPremise). I have 3 master, 1 edge and 5 workers.

I am not able to up 2 of 5 NodeManager on workers. 3 of them are ok, and 2 of them give me a follow log (attach), without error but a Warning with "NullPointerException":.

When I put the NodeManager to run, on Cloudera Manager it doens't fail, but I got two alerts, as follow:

- NodeManager can not connect to ResourceManager

- ResourceManager could not connect to Web Server of NodeManager

Also, I can't access the /jmx of the server. And, when I run NodeManager by Cloudera Manager, my CPU going to use of 100%.

On that 2 workers, I have RegionServer and DataNode working fine, the problem is only with NodeManager.

Please, any suggest?

[#2 · Accepted solution]

Hi guys!

Finally we solved the problem. To fix it, we moved all content from  "yarn.nodemanager.recovery.dir" config path to another one (i.e mv yarn-rm-recovery yarn-rm-recovery-backup) and we created yarn-rm-recovery again, grant permisison to yarn:hadoop to folder.

After that, we can start NodeManager with no error.

Thanks all!

[#3]

That is great, thank you for sharing the solution! 

Best regards

 Miklos