Important Details
How are you running Sentry?
On-Premise Sentry without Docker
version 9.1.1
Description
We are using on-premise Sentry 9.1.1 without docker. And recently i find that the sentry uWSGI wokers often consume too much CPU time(100% or more). The server is 8C16G. The web server is configured as default, which starts one master uWSGI woker and 3 sub uWSGI workers. When all the uWSGI workers consume 100% CPU, operation on Sentry frontend webpage will be blocked.
As below, the Sentry uWSGI progress keeps consuming 100% CPU for very long time. And i checked the log of sentry run web, there are very few http request logs. So it’s not caused by too many concurrent requests. And i also checked the events queues, all the queues are empty. So it’s not caused by too many backlog events. So i wonder what causes the uWSGI’s high usage of CPU.
Steps to Reproduce
What you expected to happen
The progresses consume low %CPU.
Possible Solution
As i have mentioned above, i have checked the log of sentry run web, there are few logs of http requests. So it’s not caused by too many concurrent requests. And all the event queues are empty, which means it is not caused by too many backlog events.