In a discussion titled “Aurora I/O” under the Hosting category, pqMMA inquired about running Discourse on AWS EC2 with Aurora RDS, noting unusually high I/O costs—5.5 billion monthly requests—for a forum with ~1,000 active users. Falco, who runs this forum on EC2/RDS, questioned why Aurora was chosen over standard PostgreSQL RDS. pqMMA admitted uncertainty but shared their developer’s rationale: Aurora was selected for dynamic storage provisioning and easier management. Falco recommended using RDS Performance Insights to investigate. In response to Falco, pqMMA acknowledged the trade-off—avoiding disk space management at higher cost—and noted that Aurora previously failed some Discourse compatibility tests. Falco confirmed their team uses PostgreSQL RDS across hundreds of Discourse instances. pqMMA expressed intent to explore migration, and later confirmed they successfully migrated from Aurora to standard PostgreSQL, citing Falco’s feedback as instrumental.
I am curious if anyone else has Discourse running on AWS through an EC2 instance and RDS setup?
We have run into an issue where the amount of I/O requests is through the roof and creating significant cost. We are currently processing 5.5 billion monthly I/O requests, which seems like a lot for a forum of maybe a 1,000 active users.
That is a good question. I’m not sure why that decision was made in retrospect. I assume the cost would be cheaper for standard PostgreSQL, but also trying to determine if there is a setup issue causing more I/O requests than there should be.
It’s part of the trade-off. No need to handle disk space alerts in exchange for paying more for the service.
I tested Aurora (couple of years ago) and it didn’t pass all Discourse tests at the time. We run hundreds of Discourse instances in AWS, but we use PostgreSQL RDS in all of those.