Files
awoooi/scripts/reboot-recovery/repair-110-startup-script-stub.sh
Your Name c1bc1666f8
Some checks failed
CD Pipeline / workflow-shape (push) Successful in 0s
CD Pipeline / cancel-stale-cd (push) Has been skipped
CD Pipeline / tests (push) Failing after 1m45s
CD Pipeline / build-and-deploy (push) Has been skipped
CD Pipeline / post-deploy-checks (push) Has been skipped
AI 技術雷達監控 / ai-technology-watch (push) Successful in 34s
AWOOOI Harbor 110 Local Repair / workflow-shape (push) Successful in 0s
AWOOOI Harbor 110 Local Repair / harbor-110-local-repair (push) Failing after 1m20s
fix(recovery): restore 110 startup script guard
2026-07-01 19:33:31 +08:00

200 lines
6.2 KiB
Bash
Executable File

#!/usr/bin/env bash
# Restore the real 110 startup recovery script when a runner fail-closed stub
# was accidentally installed at /usr/local/bin/awoooi-startup-110.sh.
#
# This script does not reboot, restart Docker, reload Nginx, touch K3s, DBs,
# firewall, secrets, .env files, sessions, SQLite, or runner registration data.
set -euo pipefail
MODE="verify"
HOST="${HOST:-wooo@192.168.0.110}"
REMOTE_PATH="${REMOTE_PATH:-/usr/local/bin/awoooi-startup-110.sh}"
RESET_FAILED="${RESET_FAILED:-1}"
START_SERVICE=0
SSH_CONNECT_TIMEOUT_SECONDS="${SSH_CONNECT_TIMEOUT_SECONDS:-6}"
SSH_STRICT_HOST_KEY_CHECKING="${SSH_STRICT_HOST_KEY_CHECKING:-accept-new}"
BACKUP_SUFFIX="${BACKUP_SUFFIX:-before-full-startup-restore-$(date +%Y%m%d-%H%M%S)}"
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
LOCAL_SCRIPT="$SCRIPT_DIR/awoooi-startup-110.sh"
REMOTE_TMP="/tmp/awoooi-startup-110.sh.restore-${BACKUP_SUFFIX}"
SSH_OPTS=(
-o BatchMode=yes
-o ConnectTimeout="$SSH_CONNECT_TIMEOUT_SECONDS"
-o ConnectionAttempts=1
-o ServerAliveInterval=5
-o ServerAliveCountMax=1
-o StrictHostKeyChecking="$SSH_STRICT_HOST_KEY_CHECKING"
)
usage() {
cat <<'USAGE'
Usage: repair-110-startup-script-stub.sh [--verify|--dry-run|--apply] [--start-service]
Restores /usr/local/bin/awoooi-startup-110.sh from repo source if the live host
has the runner/CD-lane fail-closed stub installed in its place.
Options:
--verify Read live hash, stub marker, and service state. Default.
--dry-run Print the controlled apply plan without writing host state.
--apply Copy the real startup script, backup the live file, install,
daemon-reload, and reset only awoooi-startup-110.service.
--start-service With --apply only, start awoooi-startup-110.service after
installing. This can run docker compose up checks, so it is
intentionally opt-in.
-h, --help Show this help.
USAGE
}
while [ "$#" -gt 0 ]; do
case "$1" in
--verify)
MODE="verify"
;;
--dry-run)
MODE="dry-run"
;;
--apply)
MODE="apply"
;;
--start-service)
START_SERVICE=1
;;
-h|--help)
usage
exit 0
;;
*)
printf 'Unknown argument: %s\n' "$1" >&2
usage >&2
exit 2
;;
esac
shift
done
if [ ! -s "$LOCAL_SCRIPT" ]; then
echo "LOCAL_STARTUP_SCRIPT_MISSING path=$LOCAL_SCRIPT" >&2
exit 66
fi
local_sha256() {
if command -v sha256sum >/dev/null 2>&1; then
sha256sum "$LOCAL_SCRIPT" | awk '{print $1}'
else
shasum -a 256 "$LOCAL_SCRIPT" | awk '{print $1}'
fi
}
quote() {
printf '%q' "$1"
}
remote_readback() {
local expected_hash="$1"
local remote_path_q
remote_path_q="$(quote "$REMOTE_PATH")"
ssh "${SSH_OPTS[@]}" "$HOST" "EXPECTED_HASH=$(quote "$expected_hash") REMOTE_PATH=$remote_path_q bash -s" <<'REMOTE'
set -uo pipefail
echo "STARTUP_STUB_REPAIR_READBACK=1"
echo "remote_path=$REMOTE_PATH"
if [ -e "$REMOTE_PATH" ]; then
echo "remote_file_present=1"
echo "remote_hash=$(sha256sum "$REMOTE_PATH" 2>/dev/null | awk '{print $1}' || true)"
echo "remote_matches_expected=$([ "$(sha256sum "$REMOTE_PATH" 2>/dev/null | awk '{print $1}' || true)" = "$EXPECTED_HASH" ] && echo 1 || echo 0)"
if head -20 "$REMOTE_PATH" 2>/dev/null | grep -Fq "runner/CD lane is fail-closed"; then
echo "remote_is_runner_failclosed_stub=1"
else
echo "remote_is_runner_failclosed_stub=0"
fi
file "$REMOTE_PATH" 2>/dev/null | sed 's/^/remote_file_type=/'
else
echo "remote_file_present=0"
echo "remote_hash="
echo "remote_matches_expected=0"
echo "remote_is_runner_failclosed_stub=0"
fi
echo "service_enabled=$(systemctl is-enabled awoooi-startup-110.service 2>/dev/null || true)"
echo "service_active=$(systemctl is-active awoooi-startup-110.service 2>/dev/null || true)"
echo "service_result=$(systemctl show awoooi-startup-110.service -p Result --value 2>/dev/null || true)"
echo "service_exec_status=$(systemctl show awoooi-startup-110.service -p ExecMainStatus --value 2>/dev/null || true)"
echo "system_failed_units=$(systemctl --failed --no-legend --plain 2>/dev/null | wc -l | tr -d ' ')"
REMOTE
}
print_plan() {
cat <<PLAN
AWOOOI_110_STARTUP_STUB_REPAIR_PLAN=1
mode=$MODE
host=$HOST
remote_path=$REMOTE_PATH
local_script=$LOCAL_SCRIPT
local_hash=$(local_sha256)
backup_path=${REMOTE_PATH}.${BACKUP_SUFFIX}
reset_failed=$RESET_FAILED
start_service=$START_SERVICE
forbidden_actions=reboot,docker_restart,nginx_reload,k3s_change,db_change,firewall_change,secret_read,runner_registration_read
rollback_hint=restore ${REMOTE_PATH}.${BACKUP_SUFFIX} to $REMOTE_PATH and systemctl daemon-reload
PLAN
}
apply_restore() {
local expected_hash
local start_command=":"
expected_hash="$(local_sha256)"
scp "${SSH_OPTS[@]}" "$LOCAL_SCRIPT" "$HOST:$REMOTE_TMP" >/dev/null
if [ "$START_SERVICE" -eq 1 ]; then
start_command="sudo -n systemctl start awoooi-startup-110.service"
fi
ssh "${SSH_OPTS[@]}" "$HOST" \
"REMOTE_TMP=$(quote "$REMOTE_TMP") REMOTE_PATH=$(quote "$REMOTE_PATH") BACKUP_SUFFIX=$(quote "$BACKUP_SUFFIX") RESET_FAILED=$(quote "$RESET_FAILED") EXPECTED_HASH=$(quote "$expected_hash") bash -s" <<REMOTE
set -euo pipefail
if ! sudo -n true >/dev/null 2>&1; then
echo "BLOCKED sudo_nopasswd_required_for_startup_script_install" >&2
exit 77
fi
sudo -n chattr -i "\$REMOTE_PATH" "\$(dirname "\$REMOTE_PATH")" >/dev/null 2>&1 || true
if [ -e "\$REMOTE_PATH" ]; then
sudo -n cp -p "\$REMOTE_PATH" "\$REMOTE_PATH.\$BACKUP_SUFFIX"
fi
sudo -n install -o root -g root -m 0755 "\$REMOTE_TMP" "\$REMOTE_PATH"
remote_hash=\$(sha256sum "\$REMOTE_PATH" | awk '{print \$1}')
if [ "\$remote_hash" != "\$EXPECTED_HASH" ]; then
echo "BLOCKED installed_hash_mismatch expected=\$EXPECTED_HASH actual=\$remote_hash" >&2
exit 78
fi
sudo -n systemctl daemon-reload
if [ "\$RESET_FAILED" = "1" ]; then
sudo -n systemctl reset-failed awoooi-startup-110.service >/dev/null 2>&1 || true
fi
rm -f "\$REMOTE_TMP"
$start_command
REMOTE
remote_readback "$expected_hash"
}
print_plan
case "$MODE" in
dry-run)
echo "DRY_RUN=1"
exit 0
;;
verify)
remote_readback "$(local_sha256)"
;;
apply)
apply_restore
;;
*)
echo "Unsupported mode: $MODE" >&2
exit 2
;;
esac