Question

I’ve been looking at MongoDB. Feels good.

I added some indexes to a collection, uploaded a bunch of data, then removed all the data, and I noticed the indexes did not change size, similar to the behavior reported here.

If I call

db.repairDatabase()

the indexes are then squashed to near-zero. Similarly if I don’t remove all the data, but call repairDatabase(), the indexes are squashed somewhat (perhaps because unused extends are truncated?). I am getting index size from “totalIndexSize” of db.collection.stats().

However, that takes a long time (I’ve read it could be hours on a large database). It’s unclear to me how available the database is for reads or writes while it is running. I am guessing not so available.

Since I want to run as few instances of mongod as possible, I want to understand more about how indexes are managed after deletes. Can anyone point me to anything or give any advice?

Answer 1

To summarize David’s linked question:

• No way except db.repairDatabase().
• If you need to minimize downtime, set up a master/slave configuration, and “repairDatabase” the slave, then switch slave and master.