The performance of sentry

I use sentry running (with Python method) on a server (with 4 core E5 2680 v3 CPU, 8G RAM) for company project , and it’s running out of CPU (always higher than 90% , load ave 8 , mostly it’s celery and uwsgi’s process. but not stucked), it runing almost with default settings .

I tried some advice in on my local test environment(centos 7 on vmware, macOS), it seem the CPU used higher as before when server receive new message in a high speed (I test with 20message/second) . And even if I stop sending new message, redis also have many key left (after running about two hour, I got 70k keys, checked by DBSIZE), and celery process running in 90% cpu as before.

So my question is , it’s that normal sentry running with so much CPU time , or maybe I should upgrade my server’s hardware ? And are there any advice to reduce the CPU time sentry used?

Our on-premise sentry also faced the same issue. I asked before but got no reply: Sentry worker performance.
Sentry just eats too much cpu with low throughput.