If you run it after the restore on the original DB then it's *possible* that there were changes, right? (Unless this was on your laptop that on one else connects to :) ).
Here is what you can do next to try to figure out *what* the differences are (we can't really figure out the "why" until we know the "what").
In the original server and in your restored copy run this command on the database in question:
> db.db.printCollectionStats()
This will print this sort of output for each collection in the DB:
out
{
"ns" : "new.out",
"count" : 10,
"size" : 360,
"avgObjSize" : 36,
"storageSize" : 8192,
"numExtents" : 1,
"nindexes" : 1,
"lastExtentSize" : 8192,
"paddingFactor" : 1,
"systemFlags" : 1,
"userFlags" : 0,
"totalIndexSize" : 8176,
"indexSizes" : {
"_id_" : 8176
},
"ok" : 1
}
You can compare collection by collection and when you find mis-matches jot them down then we can try to figure out how they came about.
Asya
P.S. if you are going to try to duplicate the problem, I would recommend doing db.stats() on the original server BEFORE doing the dump.
Then do the dump, restore and do db.stats() in the restored database.