Just wanted to let everyone know my experience with MongoDB Atlas and see if anyone else has the same experience. I’ve been running my DB through MLab for the last couple years on their shared plan. This has been fine while my total DB size was around 1GB (3GB total counting the replica set), but with some new added features to the app the DB has grown to about 2GB (6GB total) which was causing me to have to compact the DB often to keep it under the 8GB limit. The next tier on MLab was $180/mo, over twice what I was paying, plus they still were not on mongo 3.4. I’d hears some good things about atlas, where I could get 40GB total space for $.1/hr for a 3 replica set, which was about equal to my MLab bill. Moving over was pretty easy using their built-in tools. Relaunched all my microservices with the new DB without a hitch.
The problems started almost immediately though. The first day I was hit with a $5 “AWS Data Transfer” fee. Atlas claimed that I used over 66GB of data on that day alone at $.09/GB. I never saw this kind of data usage with my app before. The app itself consists of 5 microservices and a handful of client connections. It does update data on a minute-by-minute schedule, but only a handful of items. The next day was worse. Atlas reported that I had used 99GB of data transfer for $9. The next 129GB and the next four days it averaged in between 99-120GB a day. The first week I had wracked up $50 in data use charges. At that rate my use would have been double or triple a dedicated plan on MLab was. I asked Atlas about it but they said it was my app was generating that data.
Out of desperation I created my own 3 member replica set on Digital Ocean using their $20/mo server (3X for $60 a month) which was a lot easier than I though to set up (in retrospect I should have done this sooner). Same level of security as my Atlas deployment. Moved my DB over. Atlas reported a max thruput on the database at over 10MB/s, with an average of around 3.5MB a second (as reported using MongoDB Compass). On my DO deployment the max as like 500KB/s, with an average of only 61KB/s. Many factors lower than Atlas was reporting and more inline with what MLab was reporting.
For now I’m going to stick with my own deployment. Much cheaper and backups don’t cost and arm and leg. It doesn’t have all the fancy reporting tools that Atlas has, but performance is the same or better.
Has anyone else seen these kind of data transfer fees on Atlas. I want to recommend it to other clients, as I like that it is on a current version of Mongo and their toolset is better than MLab, but can’t in good conscience do it if data usage fees will kill them.