MTV-1343 | Add pprof to the controller

cmd/forklift-controller/main.go

@@ -18,6 +18,7 @@ package main
 
 import (
 	"net/http"
+	_ "net/http/pprof"
 	"os"
 	"time"
 
@@ -69,7 +70,20 @@ func main() {
 	// Start prometheus metrics HTTP handler
 	log.Info("setting up prometheus endpoint :2112/metrics")
 	http.Handle("/metrics", promhttp.Handler())
-	go http.ListenAndServe(":2112", nil)
+	go func() {
+		err := http.ListenAndServe(":2112", nil)
+		if err != nil {
+			log.Info("failed to setup the metrics endpoint")
+		}
+	}()
+
+	log.Info("setting up profiling endpoint :6060")
+	go func() {
+		err := http.ListenAndServe("localhost:6060", nil)
+		if err != nil {
+			log.Info("failed to setup the profiling endpoint")
+		}
+	}()
 
 	// Get a config to talk to the apiserver
 	log.Info("setting up client for manager")

docs/profiling-guide.md

@@ -0,0 +1,31 @@
+# Profiling of Forklift
+This is profiling doc for the forklift which contains few examples. It is not full documentaiton on how to use the
+pprof, for that please visit the [pprof](https://pkg.go.dev/net/http/pprof).
+
+### Forward the pods profiling port
+The pprof is running within the inventory and controller container. The port is not exposed to so you need to forward
+the port to your local machine with the following command:
+
+`oc port-forward -n konveyor-forklift pods/forklift-controller-5c99c54dd8-77blm 6060:6060`
+
+### Get running goroutines and heap
+You can get the running goroutines and the heap allocations. This data can be used to locate possible memory leaks.
+Using the option `-http=:9001` the pprof tool will open a webserver for the analysis.
+
+`go tool pprof -http=:9001 localhost:6060/debug/pprof/goroutine`
+
+`go tool pprof -http=:9002 localhost:6060/debug/pprof/heap`
+
+### Get trace over 5s
+You can get all calls that the server did and how long they took within x seconds using the trace. 
+
+`curl -k http://localhost:6060/debug/pprof/trace\?debug\=1\&seconds\=5 -o cpu-trace.out`
+
+This will create a `cpu-trace.out` file which you open within the webserver using command:
+
+`go tool trace -http=:9003 ./cpu-trace.out`
+
+### List all routines
+Additionally, if you want a list of goroutines you can query it usign commnad:
+
+`curl -k -o stage.out http://localhost:6060/debug/pprof/goroutine\?debug\=1`