fix(ops): bound CD DB probes and restore SigNoz collector

scripts/backup/backup-signoz.sh

@@ -1,10 +1,11 @@
 #!/bin/bash
 # =============================================================================
 # WOOO AIOps - SignOz 備份腳本 (ClickHouse + SQLite)
-# 版本: 1.1.0
+# 版本: 1.2.0
 # 建立日期: 2026-04-05
 # 2026-04-05 Claude Code: 新增 SignOz 分散式追蹤備份 — 首席架構師備份審計
 # 2026-04-05 Claude Code: v1.1 修正 tar pipeline exit code 處理 + || true
+# 2026-07-15 Codex: v1.2 確保任何 exit/signal/failure 都會受控還原 OTEL Collector
 # =============================================================================
 
 set -euo pipefail
@@ -14,11 +15,42 @@ source "$(dirname "$0")/common.sh"
 SERVICE="signoz"
 LOCAL_REPO="${BACKUP_BASE}/signoz"
 DUMP_DIR="/tmp/signoz-backup-$$"
+COLLECTOR_RESTORE_PENDING=0
+CLEANUP_COMPLETE=0
+
+restore_collector() {
+    if [ "${COLLECTOR_RESTORE_PENDING}" -ne 1 ]; then
+        return 0
+    fi
+
+    log_info "重啟 signoz-otel-collector..."
+    if docker start signoz-otel-collector 2>/dev/null; then
+        COLLECTOR_RESTORE_PENDING=0
+        return 0
+    fi
+
+    log_warn "signoz-otel-collector 重啟失敗"
+    return 1
+}
 
 cleanup() {
-    # 確保 collector 已重啟
-    docker start signoz-otel-collector 2>/dev/null || true
+    local exit_code=$?
+
+    if [ "${CLEANUP_COMPLETE}" -eq 1 ]; then
+        return "${exit_code}"
+    fi
+    CLEANUP_COMPLETE=1
+
+    # cleanup 期間忽略重複 signal，避免還原與刪除流程重入。
+    trap '' HUP INT TERM
+    restore_collector || true
     rm -rf "${DUMP_DIR}"
+    return "${exit_code}"
+}
+
+on_signal() {
+    local signal_number="$1"
+    exit "$((128 + signal_number))"
 }
 
 backup_volume() {
@@ -44,6 +76,12 @@ backup_volume() {
 
 main() {
     local start_time=$(date +%s)
+
+    trap cleanup EXIT
+    trap 'on_signal 1' HUP
+    trap 'on_signal 2' INT
+    trap 'on_signal 15' TERM
+
     log_info "========== 開始 SignOz 備份 =========="
     mkdir -p "${DUMP_DIR}"
 
@@ -51,13 +89,14 @@ main() {
 
     # Step 1: 停止 OTEL Collector 確保數據一致性
     log_info "暫停 signoz-otel-collector 以確保數據一致性..."
+    # 先 arm 還原再停止，消除 signal 發生在 stop 與狀態記錄之間的空窗。
+    COLLECTOR_RESTORE_PENDING=1
     docker stop signoz-otel-collector 2>/dev/null || log_warn "signoz-otel-collector 未在運行，繼續"
     docker stop signoz-telemetrystore-migrator 2>/dev/null || true
 
     # Step 2: 備份 ClickHouse volume (排除 tmp 目錄降低體積)
     backup_volume "signoz-clickhouse" "${DUMP_DIR}/clickhouse_${timestamp}.tar.gz" "--exclude=/data/tmp" || {
         log_error "ClickHouse volume 備份失敗"
-        cleanup
         notify_clawbot "failed" "${SERVICE}" "SignOz ClickHouse 備份失敗"
         exit 1
     }
@@ -68,15 +107,14 @@ main() {
     }
 
     # Step 4: 重啟 Collector
-    log_info "重啟 signoz-otel-collector..."
-    docker start signoz-otel-collector 2>/dev/null || log_warn "signoz-otel-collector 重啟失敗"
+    # 若此處失敗，保留 pending 讓 EXIT cleanup 再做一次有界還原。
+    restore_collector || true
 
     # Step 5: 初始化 Restic 倉庫
     if [ ! -d "${LOCAL_REPO}/data" ]; then
         log_info "初始化 Restic 倉庫: ${LOCAL_REPO}"
         restic -r "${LOCAL_REPO}" init --password-file "${RESTIC_PASSWORD_FILE}" 2>&1 || {
             log_error "Restic 倉庫初始化失敗"
-            rm -rf "${DUMP_DIR}"
             exit 1
         }
     fi
@@ -92,8 +130,6 @@ main() {
     # Step 7: GFS 清理
     cleanup_old_backups "${LOCAL_REPO}"
 
-    rm -rf "${DUMP_DIR}"
-
     local end_time=$(date +%s)
     local duration=$((end_time - start_time))
     log_success "========== SignOz 備份完成 (${duration}s) =========="