Application metrics can be enabled to see your application's metrics.
Devtron provides certain metrics (CPU and Memory utilization) for each application by default i.e. you do not need to enable “Application metrics”. However, prometheus needs to be present in the cluster and the endpoint of the same should be updated in Global Configurations --> Clusters & Environments section.
There are certain advanced metrics (like Latency, Throughput, 4xx, 5xx, 2xx) which are only available when "Application metrics" is enabled from the Deployment Template. When you enable these advanced metrics, devtron attaches a envoy sidecar container to your main container which runs as a transparent proxy and passes each request through it to measure the advanced metrics.
Note: Since, all the requests are passed through envoy, any misconfiguration in envoy configs can bring your application down, so please test the configurations in a non-production environment extensively.
CPU usage is a utilization metric that shows the overall utilization of cpu by an application. It is available as both, aggregated or per pod.
Memory usage is a utilization metric that shows the overall utilization of memory by an application. It is available as both, aggregated or per pod.
This application metrics indicates the number of request processed by an application per minute.
This metrics indicates the application’s response to client’s request with a specific status code i.e 1xx(Communicate transfer protocol-level information), 2xx(Client’s request was accepted successfully), 3xx(Client must take some additional action to complete their request), 4xx(Client side error) or 5xx(Server side error).
Latency metrics shows the latency for an application. Latency measures the delay between an action and a response.
99.9th percentile latency: The maximum latency, in seconds, for the fastest 99.9% of requests.
99th percentile latency: The maximum latency, in seconds, for the fastest 99% of requests.
95th percentile latency: The maximum latency, in seconds, for the fastest 95% of requests.
Note: We also support custom percentile input inside the dropdown .A latency measurement based on a single request is not meaningful.