#!/bin/bash

# 清理 Kubernetes 集群中失败和未就绪的 Pod
# tags: kubernetes, k3s, cleanup, maintenance
# description: 自动清理所有失败、未就绪或处于错误状态的 Pod
# title: Kubernetes Pod 清理脚本
# createdAt: 2025-11-26

set -e

# 颜色定义
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
RED='\033[0;31m'
NC='\033[0m' # No Color

echo "======================================"
echo "Kubernetes Pod 清理脚本"
echo "======================================"
echo ""

# 获取所有失败或未就绪的 Pod
echo -e "${YELLOW}正在扫描失败的 Pod...${NC}"
echo ""

# 查找所有问题 Pod
FAILED_PODS=$(kubectl get pods -A -o json | jq -r '
  .items[] | 
  select(
    .status.phase != "Running" and 
    .status.phase != "Succeeded" or
    (.status.containerStatuses // [] | any(.ready == false))
  ) | 
  "\(.metadata.namespace) \(.metadata.name) \(.status.phase // "Unknown")"
' 2>/dev/null || echo "")

if [ -z "$FAILED_PODS" ]; then
    echo -e "${GREEN}✓ 没有发现失败的 Pod${NC}"
    echo ""
    exit 0
fi

echo "发现以下问题 Pod："
echo "$FAILED_PODS" | while read namespace name status; do
    echo -e "${RED}  [$status] $namespace/$name${NC}"
done
echo ""

# 确认删除
read -p "是否删除这些 Pod？(y/N): " -n 1 -r
echo
if [[ ! $REPLY =~ ^[Yy]$ ]]; then
    echo -e "${YELLOW}取消清理${NC}"
    exit 0
fi

# 删除问题 Pod
echo ""
echo -e "${YELLOW}开始清理...${NC}"
DELETED=0
FAILED=0

echo "$FAILED_PODS" | while read namespace name status; do
    if [ -n "$namespace" ] && [ -n "$name" ]; then
        printf "删除 ${namespace}/${name} ... "
        if kubectl delete pod "$name" -n "$namespace" --grace-period=0 --force 2>/dev/null; then
            echo -e "${GREEN}✓${NC}"
            ((DELETED++)) || true
        else
            echo -e "${RED}✗${NC}"
            ((FAILED++)) || true
        fi
    fi
done

echo ""
echo -e "${GREEN}======================================"
echo "清理完成"
echo "======================================${NC}"
echo ""
echo "提示：相关的 Deployment/StatefulSet 会自动重新创建 Pod"
echo ""