#!/usr/bin/env bash
# scripts/dns-inflight-4snap.sh <runlabel> — STRICT/LOGICAL flip 중 in-flight(장수명 40s) 요청 1회 +
# flip(+5s) + 4점 스냅샷(BEFORE/MID1+12s/MID2+30s/AFTER end+3s)으로 "언제" 커넥션이 정리되는지 규명한다.
#
# 전제: scripts/dns-lab-setup.sh 로 랩이 떠 있고, scenarios/50-dns-resolution/20-backends.yaml 의
# `/slow`(10KB/s 제한, 409610B ≈ 40초 응답) 엔드포인트가 반영된 상태. SE(strict/logical)는 이 스크립트
# 실행 전 원하는 변형으로 미리 apply 해둘 것(예: 40-serviceentry-strict.yaml 또는 41-*-logical.yaml).
# 스냅샷은 netshoot 사이드카(실제 트래픽 지점) 기준 — fortio 아님, 하네스 dns-flip-test.sh 의
# estat() 관례와 다르니 주의.
#
# 출처: docs/test-reports/2026-07-02_104250_dns-strict-inflight.md (2점 스냅샷, in-flight 절단 미재현)
#       docs/test-reports/2026-07-02_112113_dns-inflight-part2.md (이 4점 스냅샷으로 destroy_local 타이밍 확정)
#
# usage: bash scripts/dns-inflight-4snap.sh <runlabel>
set -uo pipefail
RUN="${1:?usage: dns-inflight-4snap.sh <runlabel>}"
CTX="${CTX:-homelab}"; NS="${NS:-dns-lab}"
DIR="$(cd "$(dirname "$0")/.." && pwd)"
WORK="$DIR/tmp/dns-inflight"; mkdir -p "$WORK"

snap(){ # $1 = snapshot name
  echo "=== ${RUN} ${1} snapshot @ $(date +%H:%M:%S) ==="
  echo "-- dig --"
  kubectl --context="$CTX" -n "$NS" exec deploy/netshoot -- dig +short gslb.lab.internal 2>/dev/null
  echo "-- endpoints (netshoot) --"
  istioctl --context="$CTX" proxy-config endpoints "deploy/netshoot.$NS" --cluster "outbound|443||gslb.lab.internal" 2>/dev/null
  echo "-- cluster stats (gslb, netshoot sidecar) --"
  kubectl --context="$CTX" -n "$NS" exec deploy/netshoot -c istio-proxy -- pilot-agent request GET "stats?filter=gslb" 2>/dev/null \
    | grep -E 'cluster\.outbound\|443.*(upstream_cx_total|upstream_cx_destroy|upstream_cx_active|upstream_cx_connect_fail|upstream_rq_total|membership_change)' \
    | sed 's#cluster.outbound|443||gslb.lab.internal.##'
  echo "-- endpoint cx (clusters, netshoot sidecar) --"
  kubectl --context="$CTX" -n "$NS" exec deploy/netshoot -c istio-proxy -- pilot-agent request GET clusters 2>/dev/null \
    | grep -E 'outbound\|443\|\|gslb\.lab\.internal::[0-9]' \
    | grep -E 'cx_active|cx_total|rq_total|rq_active|health_flags' \
    | sed 's#outbound|443||gslb.lab.internal::##'
  echo
}

sleep_until(){ # $1 = target epoch
  local now; now=$(date +%s)
  local d=$(( $1 - now ))
  [ "$d" -gt 0 ] && sleep "$d"
}

{
snap BEFORE

T0=$(date +%s)
echo "CURL_LAUNCH=$(date +%H:%M:%S) (T0=$T0)"
kubectl --context="$CTX" -n "$NS" exec deploy/netshoot -- sh -c 'date +%H:%M:%S > /tmp/rX_podstart.txt; curl -s -m 90 -o /tmp/rX.bin -w "CURL_EXIT=%{exitcode} SIZE=%{size_download} TIME=%{time_total} CODE=%{http_code}"  http://gslb.lab.internal/slow; rc=$?; echo; echo "RC=$rc"; date +%H:%M:%S > /tmp/rX_podend.txt' > "$WORK/${RUN}_curl_out.txt" 2>"$WORK/${RUN}_curl_stderr.txt" &
CURLPID=$!

sleep_until $((T0+5))
echo "FLIP_ISSUED=$(date +%H:%M:%S) (t=+$(( $(date +%s) - T0 ))s)"
kubectl --context="$CTX" -n "$NS" exec deploy/lab-dns -c writer -- sh -c "printf '%s gslb.lab.internal\n' '10.250.27.55' > /hosts/addn"
echo "FLIP_DONE=$(date +%H:%M:%S) (t=+$(( $(date +%s) - T0 ))s)"
echo

sleep_until $((T0+12))
echo "[t=+$(( $(date +%s) - T0 ))s]"; snap MID1

sleep_until $((T0+30))
echo "[t=+$(( $(date +%s) - T0 ))s]"; snap MID2

wait $CURLPID
CURL_END=$(date +%s)
echo "CURL_WAIT_DONE=$(date +%H:%M:%S) (t=+$(( CURL_END - T0 ))s)"
echo "-- curl result --"
cat "$WORK/${RUN}_curl_out.txt"
echo

sleep 3
echo "[t=+$(( $(date +%s) - T0 ))s]"; snap AFTER

echo "-- received file check (netshoot) --"
kubectl --context="$CTX" -n "$NS" exec deploy/netshoot -- sh -c 'wc -c /tmp/rX.bin; head -c 15 /tmp/rX.bin; echo'
} | tee "$WORK/${RUN}_full.txt"