search

discourse / prometheus_exporter

A framework for collecting and aggregating prometheus metrics
MIT License
525 stars 153 forks source link

Confusing http_request_queue_duration_seconds results #278

Closed pdsmcgavin closed 1 year ago

pdsmcgavin commented 1 year ago

More of a heads up for anyone else who was in my position today. Amazon Application Load Balancer request tracing header is only accurate to a second so if you're wondering why your queue duration is averaging to around 0.5 seconds (and your 99th quantile is 0.99 seconds) then you'll need to add something along the lines of:

proxy_set_header X-Request-Start "t=${msec}";

To get accurate results for this. Possibly worth updating the ReadMe if this isn't already there?

(Cheers for the gem though as it's saved us a load of bother)

pdsmcgavin commented 1 year ago

image

The grafana panel that was causing some suspicion

sosedoff commented 1 year ago

@pdsmcgavin See https://github.com/discourse/prometheus_exporter/pull/279

pdsmcgavin commented 1 year ago

Awesome thank you

vitobotta commented 3 weeks ago

@pdsmcgavin Hi! What dashboard is that?

pdsmcgavin commented 2 weeks ago

@pdsmcgavin Hi! What dashboard is that?

Sorry to say that it's a dashboard I don't have any more. That panel was using ruby_http_request_queue_duration_seconds I believe though if that helps

vitobotta commented 2 weeks ago

@pdsmcgavin Hi! What dashboard is that?

Sorry to say that it's a dashboard I don't have any more. That panel was using ruby_http_request_queue_duration_seconds I believe though if that helps

gotcha, thanks!