#!/usr/bin/env bash
# T79: Pod 종료 시간은 Envoy drain(terminationDrainDuration)과 app preStop이
# 직렬(합)이 아니라 병렬(max)로 진행됨을 검증.
#
# NOTE (from investigation): the manifest's app image (hashicorp/http-echo:0.2.3)
# has NO shell, so its `sh -c "sleep 15"` preStop hook fails INSTANTLY (does not
# actually sleep 15s), and the binary does not trap SIGTERM as PID 1, so it runs
# to the full terminationGracePeriodSeconds and is SIGKILLed. This means the
# aggregate `total_termination_seconds` metric is confounded by TGPS, not a clean
# test of the max(15,20) vs sum(15+20) hypothesis. This script still reproduces
# the original spec commands AND adds per-container timestamp polling, which is
# what actually demonstrates the parallel behavior (istio-proxy exits cleanly at
# ~20s = terminationDrainDuration, independent of the app container which is only
# killed at ~60s = terminationGracePeriodSeconds).
set -euo pipefail

NS="istio-vt-t79"
WORKDIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

cleanup() {
  echo "[cleanup] deleting namespace ${NS}"
  kubectl delete namespace "${NS}" --wait=false --ignore-not-found
}
trap cleanup EXIT

echo "[setup] creating namespace ${NS}"
kubectl create namespace "${NS}"
kubectl label namespace "${NS}" istio-injection=enabled

echo "[setup] applying manifest.yaml"
kubectl apply -f "${WORKDIR}/manifest.yaml"

echo "[wait] deployment available"
kubectl -n "${NS}" wait --for=condition=available deploy/paralleltiming-echo --timeout=60s
echo "[wait] client pod ready"
kubectl -n "${NS}" wait --for=condition=Ready pod/client --timeout=60s

POD=$(kubectl -n "${NS}" get pod -l app=paralleltiming-echo -o jsonpath='{.items[0].metadata.name}')
echo "POD=${POD}"

# background poller: sample container state ~1/s during termination
(
  for _ in $(seq 1 90); do
    TS=$(date -u +%Y-%m-%dT%H:%M:%S.%NZ)
    LINE=$(kubectl -n "${NS}" get pod "${POD}" -o jsonpath='{range .status.containerStatuses[*]}{.name}={.state};{end}' 2>&1) || true
    if echo "${LINE}" | grep -qi "not found\|NotFound"; then
      echo "${TS} GONE"
      break
    fi
    echo "${TS} ${LINE}"
    sleep 1
  done
) > "${WORKDIR}/poll-run.log" 2>&1 &
POLLER_PID=$!

T0=$(date +%s)
kubectl -n "${NS}" delete pod "${POD}" --wait=true
T1=$(date +%s)
echo "total_termination_seconds=$((T1-T0))"

wait "${POLLER_PID}" 2>/dev/null || true
echo "[poll log] ${WORKDIR}/poll-run.log"
tail -5 "${WORKDIR}/poll-run.log" || true

echo "[done] see comments at top of this script for interpretation of the result."