I'm sorry I'm losing it.
Running RS on a machine where DN isn't running?
So then the RS can't store its regions locally. Not sure if that would ever be
a good idea or recommended.
Thought the initial question is running ZK on the same node as a RS which isn't
a good idea and a recipe for failure....
Following KISS is a much better way of life than taking Crystal Meth. Its one
way to avoid those nasty 'dead hooker problems'. *
*<rant>
<explanation>
Just to explain KISS and what I mean by a 'dead hooker'
problem...
KISS = Keep It Simple Stupid
This is an engineering principle used to teach engineering
students that the best solutions are the ones that are straight forward and
that if you attempt to get too clever, you always get some sort
of blow back in your face. It usually hurts and its always self inflicted.
'dead hooker problems' - are the theoretical problems of how
to get rid of the dead hooker from your hotel room after your party of Hookers,
Booze and either Crystal Meth or Cocaine goes terribly
wrong and you wake up the next morning with a nasty hangover and a dead body
that you have to get out of your hotel room before the cleaning ladies come
knocking on your hotel room door. While I've never experienced this... I
can't recall how many movies have this as a plot or sub plot.
Not that I'm attempting to advocate drugs or killing hookers,
unless its with a type writer or text editor when you want to write your next
failed movie script.
</explanation>
So here's my rant...
I'm not picking on the OP, but in general there's a class of posts
where the OP starts a thread by ignoring the common wisdom captured in books,
blogs and Apache wikis when setting up a cluster.
When things don't work, they ultimately post here and wonder why they
don't work.
The key to happiness is to not ignore the conventional wisdom and when
starting out with Hadoop, follow the suggested set ups. Remember that the key
is to first grok Hadoop before you attempt and doing more advanced things in
terms of cluster configurations. That is what is meant by KISS. Accept that
Hadoop is just a tool used by many to solve problems requiring a
parallel framework.
Dead Hooker problems may be a great plot device, but in real life, when
under a time crunch, they are something one should avoid. ;-)
</rant>
For those of you who don't appreciate my sense of humor, try another example...
(Also note... I don't know how this will translate to another language other
than English so the meaning of this could be lost in translation...)
Your wife has invited a bunch of her co-workers, including her boss, over for a
dinner. You, being the good spouse are responsible for some of the meal prep.
Rather than go with a tried and true recipe, you decide to try something new.
And not only try a new recipe, you also decide to improvise and try new
ingredients and do your own thing. Not really a good idea, and unless you are
incredibly lucky, or a really good cook with a talent for creating new recipes,
you are more than likely going to end up in the dog house.
Take it from a guy who usually lives in the dog house for one reason or
another... following the recipes and not trying something new when the pressure
for success is on... much less stress in your life. :-)
Again, with respect to Hadoop, there are a lot of moving parts where things can
go wrong. I've got this drinking buddy named Murphy... you know the guy, he
wrote this law... ;-)
HTH
-Mikey
On Jul 1, 2012, at 7:41 PM, Andrew Purtell wrote:
> A typical and recommended configuration is HBase RegionServer and HDFS
> DataNode colocated on the nodes. The DataNode will use locally
> attached disk to store and serve blocks.
