[SPARK-16352] Spark daemon run.sh

[zasdfgbnm]

What changes were proposed in this pull request?

I add some shell scripts to support running master and slave foreground, which makes it more convenient to write systemd service and run on HPC.

How was this patch tested?

I test it manually and it works well on my computer

[AmplabJenkins]

Can one of the admins verify this patch?

[HyukjinKwon]

(I think It might be nicer if this contribution follows https://cwiki.apache.org/confluence/display/SPARK/Contributing+to+Spark)

[zasdfgbnm]

I see similar pull request here:
#3881

But I didn't get what happened to that and why it was closed without adding any support for running in foreground.

[srowen]

Please don't open a duplicate JIRA. You can see the discussion on the previous issue and JIRA, which is WontFix, mostly because it ultimately wasn't followed up on. It's not out of the quesiton but you should comment on the JIRA with details about why you think it's important first.

[zasdfgbnm]

I followed that PR (#3881)

"I was under the impression all the shell scripts were getting refactored and that this patch had become obsolete. I agree it's best to close this out." by @hellertime

I'm not sure what happened to that. It seems that the previous PR was closed because the scripts had changed a lot before that PR was merged.

I didn't see anything that support running in foreground and nobody mentioned in that issue that running in foreground is useless and will never be supported.

"This patch has mostly gone stale at this point and I would recommend that we close it for now. If you prefer feel free to open a new updated one." by @andrewor14
I would say, this is the updated one.

[srowen]

I'm suggesting you keep the conversation together with the original JIRA, at least. I don't know if it was going to be merged or not, but it was not updated by the author in any event.

[zasdfgbnm]

OK, but will the status of the original JIRA keep being "won't fix", if someone decide to reopen the discussion on it?
Can anyone give a comment on whether the running on foreground feature are welcome in spark upstream?
Also, if there has already been a good solution to run in foreground easily enough (so that my PR will be useless), could anyone send me a link to the document describing this?