HBase regions are moved on the same machine

[#1]

Hello,

I have several HBase tables (each with a bunch of regions), and I do not understand why, but Hbase tends to put all the regions of the same table on the same machine.

However, I already moved the regions in order to have regions from each table on all workers (with the move command), and the day after I saw that the regions were again on either less machines, either on one machine. 

I tried to do 'balance_switch false' but with no success (the regions were still moved on one machine..).

I do not want to have them on less machines, because if all the regions are on one machine, then only one machine is working...

Do I get it wrong? Should I let Hbase move the regions on one machine/less machines?

Thank you!

[#2]

Hello, I also have a quick question.

On the HBase interface http://<hbase ip>:60010/table.jsp?name=table_name we can see where is each region located.

However, normally I have a replication of 3.
How is this done? For a region X, is the hole region duplicated on 2 other servers? How can I found out the 2 other servers on wich I can find this X region?

[#3 · Accepted solution]

I found a partial answer to my question:

http://hadoop-hbase.blogspot.fr/2013/07/hbase-and-data-locality.html 

Since I am in DEV, I start and Stop the machines every day. Also, we do restart the service from time to time.