AWS托管更新覆盖本地文件，如何保留数据并启用更新？

问题 #

我用AWS Beanstalk托管了一个应用，应用会把配置存在.property文件，历史数据存在.csv文件，原本运行正常。某天突然发现所有数据被回滚到初始状态：应用能正常运行，但.property里的配置和.csv的历史数据全没了。

查看Beanstalk实例历史，发现AWS执行了系统更新操作，推测是维护时用我上次上传的应用部署.zip包重新部署了应用。

我知道可以用数据库持久化数据，但改动太大。现在暂时关闭了Managed updates功能，但考虑到更新补丁，实例会逐渐过时。

请问怎么在开启托管更新的前提下，不用上次的.zip包部署，而是更新后用当前工作目录（推测是/app目录）部署？

AWS托管更新操作日志 #

2022-08-12 23:15:46 UTC+1000    
INFO
Environment update completed successfully.
2022-08-12 23:15:46 UTC+1000    
INFO
Successfully deployed new configuration to environment.
2022-08-12 23:14:38 UTC+1000    
INFO
Updating environment xxx-env's configuration settings.
2022-08-12 23:14:30 UTC+1000    
INFO
Environment update is starting.
2022-08-12 07:13:20 UTC+1000    
INFO
Environment health has transitioned from Info to Ok. Configuration update completed 83 seconds ago and took 11 minutes.
2022-08-12 07:12:37 UTC+1000    
INFO
Environment update completed successfully.
2022-08-12 07:12:37 UTC+1000    
INFO
Successfully deployed new configuration to environment.
2022-08-12 07:12:21 UTC+1000    
INFO
Removed instance [i-025394c6b5dc] from your environment.
2022-08-12 07:09:09 UTC+1000    
INFO
Deployment succeeded. Terminating old instances and temporary Auto Scaling group.
2022-08-12 07:07:50 UTC+1000    
INFO
Updating environment xxx-env's configuration settings.
2022-08-12 07:07:20 UTC+1000    
INFO
Waiting for post-deployment configuration to complete.
2022-08-12 07:07:16 UTC+1000    
INFO
Instance deployment completed successfully.
2022-08-12 07:06:39 UTC+1000    
INFO
Starting post-deployment configuration on new instances.
2022-08-12 07:06:11 UTC+1000    
INFO
Attached new instance(s) to the permanent auto scaling group awseb-e-wiuxjqswqj-stack-AWSEBAutoScalingGroup-AV.

2022-08-12 07:06:07 UTC+1000    
INFO
Detached new instance(s) from temporary auto scaling group awseb-e-wiuxjqswqj-immutable-stack-AWSEBAutoScalingGroup-1I2W.
2022-08-12 07:04:29 UTC+1000    
INFO
Waiting for instance(s) (i-0463ad18900a) to pass health checks. 
2022-08-12 07:03:21 UTC+1000    
INFO
Added instance [i-0463ad18900a] to your environment.
2022-08-12 07:03:21 UTC+1000    
INFO
Environment health has transitioned from Ok to Info. Configuration update in progress on 1 instance. 0 out of 1 instance completed (running for 2 minutes). All instances are in same availability zone (us-west-2c).
2022-08-12 07:00:52 UTC+1000    
INFO
Created temporary auto scaling group awseb-e-wiuxjqswqj-immutable-stack-AWSEBAutoScalingGroup-1I2.
2022-08-12 07:00:21 UTC+1000    
INFO
Immutable deployment policy enabled. Launching one instance with the new settings to verify health.
2022-08-12 07:00:13 UTC+1000    
INFO
Environment update is starting.
2022-08-12 07:00:11 UTC+1000    
INFO
Managed platform update is in-progress.

解决方案 #

1. 迁移数据到实例持久化目录 #

Beanstalk重新部署时会覆盖/var/app/current目录，但实例EBS卷上的其他目录不会被改动。操作步骤：

• 修改应用代码，将.property和.csv文件的存储路径改为/var/app/data
• 在应用根目录创建.ebextensions/01-setup-data-dir.config配置文件，内容如下：

commands:
  01_create_dir:
    command: "mkdir -p /var/app/data && chown webapp:webapp /var/app/data"

该配置会在每次部署或实例更新时自动创建目录并设置正确权限，数据会持久化在EBS卷上，不会被覆盖。

2. 切换部署策略为滚动更新 #

日志显示当前启用了不可变部署策略，该策略会创建全新实例并重新部署原始包，旧实例销毁后数据丢失。切换为滚动更新：

• 进入Beanstalk控制台，打开目标环境的「配置」>「部署」页面
• 将部署策略改为「滚动更新」，单实例环境可设置批次大小为100%
• 滚动更新会在现有实例上直接应用平台更新，不会重建实例，数据得以保留。

3. 配置托管更新不重新部署应用 #

在托管更新设置中，可指定仅更新平台配置，不重新部署应用包：

• 进入环境的「配置」>「平台更新」
• 在「托管更新」模块，将「应用部署选项」设为「仅更新平台配置，不重新部署应用」
• 此设置下，AWS仅更新操作系统、运行时补丁等，不会覆盖应用目录内的数据。

4. 可选：定期备份数据到S3 #

为防止EBS卷故障导致数据丢失，可添加定时任务同步数据到S3：

• 创建.ebextensions/02-s3-backup.config配置文件：

files:
  /etc/cron.daily/s3-backup:
    mode: "0755"
    owner: root
    group: root
    content: |
      #!/bin/bash
      aws s3 sync /var/app/data s3://your-bucket-name/app-data/ --delete

• 给Beanstalk实例的IAM角色添加s3:PutObject和s3:ListBucket权限，确保能正常同步。

内容的提问来源于stack exchange，提问作者Zhang Zhan