Get pprof Dumps (Backend)

Get pprof Dumps (Backend) #

Hit the pod’s IP directly over WARP. Needs the gcloud CLI and WARP connected — the GKE pod ranges are published WARP routes, so a pod IP is reachable from your laptop.

1. Get the pod names and IPs # 

kubectl --context=<context> \
  -n fs-apps get pods -l app=backend \
  -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.status.podIP}{"\n"}{end}'

Use the context gcloud setup generated for the environment you want. Naming it on the command line keeps the environment unambiguous — see name the context.

2. Pull a dump #

The backend serves pprof on port 8001. List what is available:

curl http://<pod-ip>:8001/debug/pprof/

Then take the profile you need:

# 30s CPU profile — for high CPU
go tool pprof http://<pod-ip>:8001/debug/pprof/profile?seconds=30

# heap snapshot — for growing memory
go tool pprof http://<pod-ip>:8001/debug/pprof/heap

# goroutine dump — for hangs and deadlocks
curl http://<pod-ip>:8001/debug/pprof/goroutine?debug=2 -o goroutines.txt

3. Read it #

go tool pprof drops you into an interactive shell:

top            # hottest functions
list <func>    # line-by-line cost
web            # call graph, needs Graphviz

Next: run the staging pod-list command, then curl http://<pod-ip>:8001/debug/pprof/.