search

pinpoint-apm / pinpoint

APM, (Application Performance Management) tool for large-scale distributed systems.
https://pinpoint-apm.gitbook.io/
Apache License 2.0
13.39k stars 3.75k forks source link

Wrong slow rate alert and slow count #7377

Open pengsili opened 3 years ago

pengsili commented 3 years ago

Hi All,

We are using pinpoint 1.8.2 to monitor our java applications. It helps a lot to find our bottleneck and monitor runtime performance.But we found that the slow request count may be something wrong in pinpoint-web. image Here you can see that the slow request count is 1 in recent 5 minutes. But in fact we can not find any request whose response time is more than 5s in above chart. And we setup a alert of slow request rate, the threshold is 30%. Sometimes we received the alert email but when we checked in pinpoint-web, we couldn`t find slow request during that time. Not sure if it is a bug and hope can get your help about this. Thanks a lot

emeroad commented 3 years ago

Is there any problem log in the collector?

pengsili commented 3 years ago

Hi @emeroad

We use pinpoint in docker. I run docker logs pinpoint-collector and didn`t find any error during that time.

emeroad commented 3 years ago

https://pinpoint-apm.github.io/pinpoint/faq.html#request-count-in-the-scatter-chart-is-different-from-the-ones-in-response-summary-chart-why-is-this

There is a problem with time granularity. See the link above.

pengsili commented 3 years ago

Hi emeroad

Thanks for your reply. i got it now why the count is different. How about the second question?We received slow rate alert but didn`t find any request more than 5s. From the doc, i know the slow request threshold(5 seconds) is canculated by code. Not all the same in different applications. So is there any possibility that our slow request threshold is less than 5s, like 4.5s?Or there is other reason? Thanks