summary refs log tree commit diff
path: root/_posts/2020-06-17-backup.md
blob: e25f496abd678d9a9b2f1b56c3fd16496812c6fc (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
---
layout: post
title: 论备份的重要性
tags: [备份]
--- 
  只有事情发生到自己头上才想到要解决<!--more-->    
  
# 起因
  今天早上发生了一件很糟糕的事情,一打开聊天软件就发现有人在说我维护的花火学园挂掉了,错误信息是无法访问源站。我觉得挺奇怪,服务器又没啥负载,我也有一段时间没登进去了,怎么服务器就挂掉了?   
  我试着用SSH连接,同样无法连接,我发现事情不太对劲,然后就登到了Vultr里看了看。结果发现我的服务器在00:00之后就像一个死人一样,CPU负载被拉成了一条直线,就那样保持0%的位置。我以为是因为莫名其妙的原因服务器关机了,然而我重启以后仍然没有解决问题。   
  登录到终端一看,`No bootable device`就这样显示在屏幕上,硬盘直接读不出来了,这下可不得了了,我赶紧去快照里看了一下,发现最后一次快照的时间在5月30日,也就是说如果没能恢复数据这十几天的所有信息都将消失!   
  
# 难以想象的垃圾服务商:Vultr
  首先我要做的事情当然是想办法先恢复服务,虽然那个快照有点早,但是先顶上再说吧……   
  想一想我的防护应该做的也没啥问题,而且一般成功入侵服务器的人也应该是删库然后留一条信息的那种,直接干死硬盘的我还真没见过。于是我开始发Ticket给Vultr,看看到底是怎么回事。   
  Vultr在我问完的4个小时后给出了最终的解决方案,把我的服务器直接重置,在上面安了新的操作系统然后给我赔了两个月的服务器费用……原文如下:   
> Hello,
>
> In the past 24 hours, we sent notification of a node failure impacting your cloud server listed above.
>
> Despite extensive efforts, our attempts to manually recover your cloud server were unsuccessful.
>
> Our engineering team is currently deploying new instances with the same operating system and IP and you will receive login details in a separate message. You may also deploy a backup or snapshot on a new instance with a new IP if you prefer.
>
> Our staff will be applying a two month account credit for the affected services shortly.
>
> Regards,   
> Bryan M.   
> Systems Administrator 

  哇,这真是太糟糕了,作为一家云服务器商就直接把客户的数据搞没了,然后就赔2个月的费用?要知道数据无价啊,就这么不负责任的吗?简直是不可思议啊!   
  不过我也没什么好办法了,也许他们不重装我还想着试试SystemRescueCD试试看能不能把整个磁盘复制出来,但是他们既然已经直接重装那就彻底没救了……QAQ   
  
# 亡羊补牢
  既然数据已经救不回来了,那我们也只能向前看,得想办法避免以后再出现这样的问题。因为我最近在期末阶段,比较忙,所以也不经常去打快照。虽然以前也出现过服务出问题的情况,像[MySQL挂了](/2020/01/05/devops.html)、CDN挂了、还有一次好像是交换机出问题了,但是无论如何数据从来没有丢失过。这一次数据都能丢了也真的是太糟糕了,要不是有快照,那就真成删库跑路了……         
  既然没时间打快照,我得想个办法搞一个自动打快照的东西。在网上搜了搜,还真有这样的脚本,于是我拿来改了改就装上去用了。
## 自动快照的脚本
```python
import requests
from requests import get
import re
import json


class __RPC:
    def __init__(self, api_key, name):
        self.api_key = api_key
        self.api_info = None
        self.name = name
        self.errors = {
            200: "Function successfully executed.",
            400: "Invalid API location. Check the URL that you are using.",
            403: "Invalid or missing API key. Check that your API key is present and matches your assigned key.",
            405: "Invalid HTTP method. Check that the method (POST|GET) matches what the documentation indicates.",
            412: "Request failed. Check the response body for a more detailed description.",
            500: "Internal server error. Try again at a later time.",
            503: "Rate limit hit. API requests are limited to an average of 2/s. Try your request again later."
        }

    def api_info_initial(self):
        self.api_info = {"snapshot/create":"POST","snapshot/destroy":"POST","snapshot/list":"GET","server/list":"GET"}

    def __getattr__(self, name):
        return eval("__RPC")(self.api_key, self.name + "/" + name)

    def __call__(self, **kwargs):
        if not self.api_info:
            self.api_info_initial()
        if self.name not in self.api_info:
            raise ValueError("The API is not exists.")

        if self.api_info[self.name] == "GET":
            res = requests.get("https://api.vultr.com/v1/" + self.name, headers={"API-Key": self.api_key},
                               params=kwargs)
        elif self.api_info[self.name] == "POST":
            res = requests.post("https://api.vultr.com/v1/" + self.name, headers={"API-Key": self.api_key}, data=kwargs)

        if res.status_code == 200:
            return res.status_code, res.text.strip()
        elif res.status_code in self.errors.keys():
            return res.status_code, self.errors.get(res.status_code)
        else:
            res.raise_for_status()


class Vultr:
    def __init__(self, api_key):
        self.api_key = api_key

    def __getattr__(self, name):
        return eval("__RPC")(self.api_key, name)

vultr = Vultr("API Key")
data = {'SUBID': '实例ID'}
status_code, resp = vultr.snapshot.create(**data)
requests.post("https://sc.ftqq.com/SCKEY.send",data ={"text":"快照已创建","desp": str(status_code)+resp})
# 删除旧快照
status_code, resp = vultr.snapshot.list()  # /v1/snapshot/list
if status_code != 200:
    print('获取快照列表失败' + str(status_code) + resp)
else:
    print('成功获取到快照列表')
    data_list = list(json.loads(resp).values())
    data_list.sort(key=lambda x: x['date_created'])  # 默认时间排序,由近到远
    data_list_del = data_list[::-1][9:]  # 取超过9个之后的快照
    for data_del in data_list_del:
        data = {'SNAPSHOTID': data_del.get('SNAPSHOTID')}
        status_code, resp = vultr.snapshot.destroy(**data)  # /v1/snapshot/destroy
        if status_code != 200:
            print('删除旧快照失败')
        else:
            print('成功删除一个旧快照')
```

  把这个脚本放到Crontab里,每天执行一次就行了。   
  
# 后记
  相信服务器厂商是完全靠不住的事情,自己还得想办法做好备份。我甚至在想,阿三把Intel和微软都占领了,会不会有一个阿三也跑到Vultr里,然后对着我的硬盘大喊“把你变成咖喱”之类的23333。   
  现在不过是权宜之计,以后还是得想办法把整个论坛下载到本地,至少能搞个数据库的差异备份啥的也行啊……