Cloudy with a chance of MongoDb

An interesting – and anonymous – article popped up this weekend on Hackernews about why you shouldn’t use MongoDb (copied over at MyNoSQL for posterity). There’s a nasty tone in that letter which must be stemming from some painful first-hand experience with the thingie. One can understand why they want to be anonymous – to draw attention on Mongo’s issues and not on their project or team. The rebuttals that came up right away are pretty funny though, especially this one containing interesting advice:

- RTFM/”use the latest version” - “Sometimes losing data doesn’t matter” - “Software has bugs” - “We experienced database corruption so keep at least 3 copies of everything you consider important” - “Backup often”.  Which means according to Mongo docs one of three recommended backup strategies:

1. LVM/EBS snapshot 2. Shutdown and backup 3. Backup through a replicated slave (note that said replication can stop without apparent reason)

Easy peasy, right?

I’ve been using all sorts of data stores and worked with tens or hundreds of million rows data sets, intensive writes – you name it. I’ve seen actual data corruption once, maybe twice in 12 years. But I have to confess that I haven’t seriously used MongoDB until now. Maybe I should.