Home / server / Questions / 787419
Accepted
Maris B.
Asked: 2016-07-02 07:05:16 +0800 CST

How to automatically update EC2 Linux instances without downtime by switching between two instances?

  • 772

In short - OS Amazon Linux or Ubuntu. One EC2 instance (or OpsWorks instance?) is running. Once a day start second instance, update it with latest security updates, install my app. Shutdown first instance. Repeat every day...

AWS documentation states:

By default, AWS OpsWorks automatically installs the latest updates during setup... We recommend that you... Create and start new instances to replace your current online instances. Then delete the current instances. The new instances will have the latest set of security patches installed during setup.

How to automate this process?

I know, I can start one micro instance to manage this process via CLI, but I would prefer to use some AWS built-in tool, so I can get notifications on failure, to add some redundancy (if CLI instance fails), everything continues to work on new CLI / AWS API versions etc.

Am I missing some elegant approach?

Update: Ability to use existing Elastic IP is preferred.

amazon-ec2 amazon-web-services aws-cli opsworks

2 Answers

  1. Best Answer

    Use EC2 instances behind an ELB.

    Upon launch, your nodes should download and install the latest security updates and do whatever other configuration is necessary to get your application running.

    As for cycling out your instances, once a day:

    1. Create a second EC2 node
    2. Wait for it to configure itself and become available
    3. Add the second node to the ELB
    4. Remove the old node from the ELB
    5. Shoot the old node in the head

    All of the above can be trivially automated using various AWS APIs, perhaps even as a Lambda job.

    • 5

  2. When you say "update it with latest security updates" what do you exactly mean? What is your threat profile and risk mitigation against what adversary? I use ufw and ensure that I do sudo apt update or sudo yum update && sudo yum upgrade && reboot in a cron set to once a month with near zero downtime. It is like 30 seconds for a reboot. It seems like what you are trying to do is not worth the time to implement.

    So my answer would be

    1. implement sane security lockdowns in terms of firewall, ports, proper user jails, proper user runtimes for your application, and in general limit what is running on the instance.

    2. setup crontab with automatic update script on a periodic basic

    This would not completely mitigate any downtime but ensures it is like 30 seconds a month.

    • -1