[SPARK-5689][Doc] Document what can be run in different YARN modes

[Sephiroth-Lin]

No description provided.

[srowen]

I don't object to this, although most of the YARN-related content here is almost all redundant, since it doesn't change between Python and Scala/Java, and this is a YARN-related doc. The docs are kind of out of order now since the explanation of SparkPi followed the python example. I am not sure this adds enough value.

[srowen]

(That is, I don't think this is quite what https://issues.apache.org/jira/browse/SPARK-5689 is getting at; a simple statement of what can be run now might be sufficient rather than duplicating this much.)

[SparkQA]

Test build #30160 has finished for PR 5490 at commit 97ba6a8.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.
• This patch does not change any dependencies.

[SparkQA]

Test build #30161 has finished for PR 5490 at commit 8ca1efe.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.
• This patch does not change any dependencies.

[SparkQA]

Test build #30163 has finished for PR 5490 at commit 908bb72.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.
• This patch does not change any dependencies.

[tgravescs]

For this I was actually thinking just making a statement about specifically what can be run in either mode. Perhaps around paragraph on line 215 where we discuss the two modes.

interactive shell(s) (both pyspark, spark-sql, spark-shell) can only be in used client mode. That might be it since we recently added the python support for cluster mode.

[tgravescs]

Jenkins, test this please

[SparkQA]

Test build #30248 has finished for PR 5490 at commit ed67797.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.
• This patch does not change any dependencies.

[andrewor14]

I actually don't think we should merge this. It's not specific to YARN that interactive shells cannot run in cluster mode. I agree that we should have a centralized table somewhere that details what applications are supported on what modes, but this doesn't seem like the right place.