efficient clean big collection

[Yulias Stolin]

I have a big collection of millions entries. I want to remove
efficiently all the entire data, but without dropping the collection.
I'm working with sharding and if i'm dropping the collection it is
automatically drop from sharding configuration.

What is the best way to do that?

db.some_coll.remove({})
takes a lot of time!!!

[A. Jesse Jiryu Davis]

Yulias, you should just drop the collection to remove all its data or, even better, keep short-lived data in separate MongoDB databases so you can drop the whole database when the data is no longer needed. (Dropping a database returns storage to the OS, unlike dropping a collection which only frees parts of the database files for Mongo to reuse.) You can write a script to recreate the indexes and sharding configuration on the next version of the collection.

[Yulias Stolin]

How can I write and store such script that each time I call it will
drop the collection, create the indexes, and insert it into shard?

[A. Jesse Jiryu Davis]

If you're developing in Python, it'd go something like:

import pymongo

connection = pymongo.Connection()

# substitute your own db and collection names

db = connection.database_name

coll = db.collection_name 

coll.drop()

# recreate whatever indexes you need ...

coll.create_index([('key', 1)])

db.command("shardcollection", "collection_name", key={'my_shard_key': 1})