What do you dislike about MongoDB Atlas?
My story with Mongo began when I started a new software position, and they had a legacy version of their software product using Atlas.
Compared to our other infrastructure bills, Mongo was significantly higher for the amount of compute and storage we used ($3K per month). This is a managed service, so you would expect to pay a premium. Ok, sure, but then I expect great functionality, performance, and support.
The main problem began with Mongo when we needed to delete some data because they tie the CPU and memory tiers to storage size, so we were overpaying. Our application would run fine off an M10 dedicated cluster (the smallest tier), but it had automatically scaled to an M50 because of storage. This is already a bit disappointing because they are forcing customers to pay for more compute and memory than they need.
So we started deleting some data, but then we ran into problems. The data deletion process was really slow and also slowed our entire cluster down, causing lag and performance issues for our end users. But hang on, this makes no sense because we are paying for more CPU and RAM than we need, so why would we have this issue?
It took us three months to delete 500GB of data. In the meantime, our bill remained the same because you can't claim the space back without compacting the database. Ok, fine. So we ran compact(), but we only freed ~100GB on the secondary clusters.
Support gave us a script to run that can see how much storage can be freed.
In the end, we had to activate an expensive additional support plan costing us $500 USD per month to get support to run a re-sync command. This should have taken their support people 10 minutes, but instead, they mucked us around going back and forth on the ticket, taking three weeks to resolve.
A year later, we needed to delete some more data. We spent another five months deleting 800GB of data. Then we ran compact() and freed 300GB. Where is our other 500GB? We contacted some humans at Mongo, who really couldn't do much other than suggest we get funding to cover the $500 support for one month. Yes, we got the $500 credit, but when I went to reactivate support, it was going to charge us for three months for one month because Mongo retroactively bills you for three months when you reactivate. Wow, we started in a bad place, now I'm beyond frustrated; this is daylight robbery.
To this day, I am still fighting to reclaim some storage, but at this point, I'm going to recommend to our CEO that our dev team put some effort into moving away completely from Mongo.
I also need to mention that Mongo recommended we use their online archive features, but when we crunched the numbers, it was still quite expensive, and we would have to do significant work to make our application work between the regular clusters and online archive. So it was significantly more logical to just put the data in AWS S3, then delete it in Mongo.
If I can summarize my experience with Mongo, and I acknowledge mine is probably quite different to most, here it is:
Overpriced for the performance you get
Sneaky billing model where they tie CPU and memory to storage
Terrible and expensive support
Sneaky extra charges on reactivating support
Bad support escalation solutions - they couldn't just turn on free 'support'
Poor database performance
Slow delete operations
Ecosystem lock-in
Forced upgrades - no LTS releases
Let me sum it up this way: if your compact() command does not free up the space that is available on your cluster, then provide the customer with free support to do so.
I hate dealing with Mongo. Nothing is simple, everything is expensive, and the performance sucks.
If you are considering using Mongo, find something else. Even if you have to take a bit more time to learn AWS Dynamo, S3, or Aurora, you should do it; you will save time and money in the long run.
Mongo, you deserve this negative review. I have given you plenty of opportunities to resolve things and have escalated issues, but you just don't care.
We wanted to move away from Mongo before; now I can't get rid of it fast enough. Review collected by and hosted on G2.com.
