RAM consumption of self-hosted sentry hovering at 22GB after upgrade

[itsafire]

Self-Hosted Version

24.11.1

CPU Architecture

x86_64

Docker Version

24.0.6

Docker Compose Version

2.28.1

Machine Specification

• My system meets the minimum system requirements of Sentry

Steps to Reproduce

We just upgraded from 23.6.2 to 24.11.1. We used to be able to conveniently run sentry on a 16GB machine on version 23.6.2 as advised in the documentation. Since the upgrade we were forced to migrate to a machine with 32GB RAM since the memory usage is now around 22GB. There are now a lot of "small" sentry (23) and snuba (20) container with each of them consume between 200MB to 300MB of RAM.

Expected Result

Since the documentation states 16GB RAM as minimum requirement we expected version 24.11.1 to work with this configuration.

Actual Result

Various container of our updated sentry OOMed with 16GB of RAM.

Installing on a clean sentry 24.11.1 without any projects will result in 13GB of RAM usage.

Installing a clean sentry with version 23.6.2 and no projects resulted in 8 GB RAM usage.

Event ID

No response

[aldy505]

Do you have high event ingestion throughput? The number of projects shouldn't be an issue, what could be the issue is the amount of event throughput.

Can you specify the list of containers that's got out of memory?

[itsafire]

We have 3.4 million transactions in 30 days. I will scale back to 16GB on the weekend to see which containers are crashing. But since system memory is completely exhausted (50MB free) OOMs are probably happening on random containers.

[aldy505]

Catching up, it's been 3 weeks. How are your instance doing? Have you scale back down to 16GB yet?