This document discusses the author's experience using MongoDB over one year, identifying things they found good, bad, and ugly. Some positives included the schema-less design, query language, and Ruby ORM support. Negatives included the 2GB database limit on 32-bit machines, migration challenges, date handling, management tools, replication support at the driver level, lack of transactions, and issues using it as a relational database. The author concluded they look forward to continued use of MongoDB despite some frustrations.
Allows Mindvalley to iterate very fast No need to mess up with MySQL migrations
This was when we were still doing RnD on the different NoSQL dbs Joffrey - somewhat like Google Analytics MV uses Linode, setup a 32 bit server, suddenly no more data after 2GB
Schemaless allows you to change your mind easily at the beginning Once you have a lot of data, migrations are a pain
Dates are stored in Unix Time (milliseconds after epoch) No niceties with a real date store like DST and Timezones Dates before v2.0 were unsigned, so having a date before 1970 resulted in a rollover
JSON - makes it really easy Almost like SQL Language-specific DSL makes it easier
Tried many: RockMongo, MongoHub, etc. Management tools are still not that mature Export/Import, dumps, export to SQL/CSV/JSON -> command line still does it better
Mongoid
Story of slaves and replica set in Oathkeeper -> it went down, it didn’t failover, turns out the driver configuration was for slaves and not for replica sets