MongoDb sharding and further extensibility

Question

MongoDB

I found an excellent guide on how to do everything sharding (something similar was even on Habré),
gliffer.ru/articles/nosql--sharding-mongodb-na-paltsah/
www.mongodb.org/display/DOCS/Sharding
blog.zawodny. com/2011/03/06/mongodb-pre-splitting-for-faster-data-loading-and-importing/

but there are a few questions left behind the scenes:

1) as I understand it, secondary keys can be sharding keys, i.e. .sharding key does not have to be unique,
for example, sharding data by regions / countries / cities

2) if we sharded data, then after some time we decided to deliver physical servers, then we can rebuild the config, reconfigure shards. That's understandable, but can I add new mongo-d instances in a way that doesn't lose data?
Example: sharding by the first letter of the alphabet:
server 1 AE
server 2 E-L
server 3 L-M
server 4 M-P
server 5 R-Y

and then it turns out that server 4 starts to overflow and you have to divide it in half:
server 4 M- About
server 6 O-P

how will data be moved in this case, which was previously located on server 4, but should now be part of server 6?

Answer 1

As far as I remember, load balancing is automatic. The subd itself will transfer data from overloaded servers to idle ones.

Answer 2

But there is one caveat!
Chunks will be stored on server 4, meaning there will be no reduction in disk space usage. You should delete chunks older than a certain time (for example, 1 day) by cron.
This functionality will be added to mongod in the next version.
The answer about removing chunks is the official 10gen recommendation.