#!/usr/bin/env bash
#
# S-16a (#122): assert the observability backplane is live against an ALREADY-RUNNING
# stack. Runs curl INSIDE the compose network (like the other verify checks) because
# the stack's published ports aren't on the CI runner's localhost — the stack is a set
# of sibling containers on the host daemon. It asks Grafana to reach its provisioned
# datasources — Prometheus via its health method, Tempo via the datasource proxy (Tempo's
# Grafana plugin implements no health method) — so it proves the datasources are wired,
# not merely that the containers started. Polls, so it tolerates a cold Grafana.
#
# Does NOT manage the stack lifecycle (the caller owns bring-up + teardown).
set -euo pipefail

TIMEOUT="${OBS_TIMEOUT:-60}"
AUTH="${GRAFANA_AUTH:-admin:admin}"

gf="$(docker ps -q --filter 'name=[-_]grafana[-_]' | head -1)"
[ -n "$gf" ] || { echo "ERROR: no running grafana container — bring the stack up first" >&2; exit 1; }
net="$(docker inspect -f '{{range $k,$_ := .NetworkSettings.Networks}}{{$k}}{{"\n"}}{{end}}' "$gf" | head -1)"
gf_ip="$(docker inspect -f '{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}' "$gf")"
base="http://$gf_ip:3000"
echo ">> grafana=$gf_ip network=$net"

# Run curl inside a throwaway container on the stack network (reaches services by IP).
net_curl() { docker run --rm --network "$net" curlimages/curl:latest "$@"; }

# poll <description> <grep -E pattern> <curl args...>
poll() {
  local desc="$1" pat="$2"; shift 2
  local deadline=$(( $(date +%s) + TIMEOUT ))
  while :; do
    if net_curl -fsS "$@" 2>/dev/null | grep -Eq "$pat"; then echo "  ✓ $desc"; return 0; fi
    if [ "$(date +%s)" -ge "$deadline" ]; then echo "  ✗ $desc ($*)" >&2; return 1; fi
    sleep 3
  done
}

echo "Checking observability backplane at $base ..."
poll "Grafana is healthy" \
  '"database":[[:space:]]*"ok"' "$base/api/health"
poll "Prometheus datasource reachable" \
  '"status":[[:space:]]*"OK"' -u "$AUTH" "$base/api/datasources/uid/prometheus/health"
poll "Tempo datasource reachable (via Grafana proxy)" \
  '"version"' -u "$AUTH" "$base/api/datasources/proxy/uid/tempo/api/status/buildinfo"
echo "Observability backplane OK."