heavy-duty-watcher/scripts/health-loop.sh

54 lines
1.6 KiB
Bash
Raw Normal View History

#!/usr/bin/env bash
# Hourly health check + self-heal for poll-loop.
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
# shellcheck source=lib/common.sh
source "${SCRIPT_DIR}/lib/common.sh"
INTERVAL_SEC="${HEALTH_INTERVAL_SEC}"
mkdir -p "${LOG_DIR}"
watcher_log "${LOG_DIR}/health.log" "health-loop started interval=${INTERVAL_SEC}s"
"${WATCHER_DIR}/scripts/health-check.sh" || true
session_name() {
if tmux has-session -t "${TMUX_SESSION}" 2>/dev/null; then
echo "${TMUX_SESSION}"
elif tmux has-session -t "${LEGACY_TMUX}" 2>/dev/null; then
echo "${LEGACY_TMUX}"
else
echo ""
fi
}
while true; do
sleep "${INTERVAL_SEC}"
set +e
"${WATCHER_DIR}/scripts/health-check.sh"
rc=$?
set -e
if ! process_running 'poll-loop\.sh'; then
watcher_log "${LOG_DIR}/health.log" "health-loop: restarting poll-loop"
sess="$(session_name)"
if [[ -n "${sess}" ]]; then
tmux list-windows -t "${sess}" -F '#{window_index} #{window_name}' \
| awk '$2=="poll-loop"{print $1}' \
| while read -r idx; do tmux kill-window -t "${sess}:${idx}" 2>/dev/null || true; done
tmux new-window -t "${sess}" -n poll-loop -c "${WATCHER_DIR}" \
"bash ${WATCHER_DIR}/scripts/poll-loop.sh"
else
tmux new-session -d -s "${TMUX_SESSION}" -n poll-loop -c "${WATCHER_DIR}" \
"bash ${WATCHER_DIR}/scripts/poll-loop.sh"
tmux new-window -t "${TMUX_SESSION}" -n health-loop -c "${WATCHER_DIR}" \
"bash ${WATCHER_DIR}/scripts/health-loop.sh"
fi
fi
if (( rc != 0 )); then
watcher_log "${LOG_DIR}/health.log" "health-loop: status non-ok rc=${rc}"
fi
done