Mongo Berlin

I've attended Mongo Berlin - a conference organized by 10gen, the makers of MongoDB. Lots of interesting talks. Here is some information besides what was shown on the official slides:

• By default, only the _id field is indexed; other indexes can be created by ensureIndex().
• Updates of one document are atomic, updates of several documents aren't.
• With the freshly brewed version 1.8.0 the maximum document size is increased from 4 MB to 16 MB.
• There are several convenient commands within the Mongo shell; e.g. use mydb is short for db = db.getSiblingDB("mydb"), and show collections is short for db.getCollectionNames().
• db.stats() shows some current statistics within the Mongo shell and db.currentOp() the currently running operation (if it takes long enough within the MongoDB server).
• Useful external commands are iostat, top etc. and mongostat.
• When journaling is used, it is a good advice to put the journal directory on separate disks.
• mongodump can be used for backups; it dumps the data into the BSON format. Use bsondump to convert BSON into JSON.
• In master/slave mode only the master node writes its own OpLog; in replica mode all nodes (i.e. the primary node and all secondary nodes) write their own OpLog.
• User authentication currently does not work in sharding mode.
• In case of a failure of the primary node within replica mode all clients get disconnected. It may take some time (5 to 30 seconds) to negotiate a new primary node; during this time clients cannot (re-)connect. rs.status() shows the current state of the replica set.