marathon perf flamegraph.md

• step 1: get the toolbox and enter it (it's a container):

wget https://raw.githubusercontent.com/mesosphere/docker-containers/master/dcos-debug/toolbox
chmod +x toolbox
./toolbox

• step 2: collect the trace (this example runs for 30s):

perf record -F 99 -p <marathon PID> -g -- sleep 30

• step 3: send the trace to us (the host machine root is accessible at /media/root from inside the toolbox)

# this puts perf.data into /tmp/perf.data on the host machine
cp perf.data /media/root/tmp/perf.data
<exit toolbox>
<email us the perf.data file>