command "yarn application" doesn't work. Getting "No HADOOP_CONF_DIR set" error.

[#1]

When I try to terminate an yarn application in CDH 5.8.x, I got the following error:

```

$ yarn application
No HADOOP_CONF_DIR set.
Please specify it either in yarn-env.sh or in the environment.

```

I've searched through all directory and found no yarn-env.sh, this error message is quite misleading.

Does some one know how to fix it?

[#2]

Please add more notes on how you've installed CDH and how you're managing it.

If you use Cloudera Manager, then does the host you use to run this command have a YARN Gateway role added and deployed on it?

Specifically, do you have a /etc/hadoop/conf/ directory with all required config files? If not this error would pop up.

[#3]

Thanks a lot! I've add YARN Gateway and yarn application is working. I must have forgot to override YARN gateway in installation.

Also, should it be a recommended setting to add YARN gateway to all nodes?

[#4]

Still has the same problem after a service reboot!

Looks like the gateway wasn't written through, strangely, my echo doesn't show any missing environment variable:

```

ubuntu@ip-172-31-23-254:~$ yarn application
No HADOOP_CONF_DIR set.
Please specify it either in yarn-env.sh or in the environment.
ubuntu@ip-172-31-23-254:~$ echo $HADOOP_CONF_DIR
/etc/spark/conf/yarn-conf:/etc/hive/conf

```

it looks like yarn get its variable elsewhere

[#5 · Accepted solution]

You do not need to set these env-vars manually. They should be set automatically for you in CDH.

The env-var, if you do choose to set it manually for some reason, must point to a singular directory of configs vs. multiple in a classpath style as it appears in your output.

Can you retry with an unset HADOOP_CONF_DIR since its handled automatically and your value's causing it to not use the automatic location? Does /etc/hadoop/conf/ exist now with usable files under it?