Search code examples
amazon-ec2amazon-web-servicesamazon-ebs

How should I approach my EC2 contingency plan?

I've browsed through other questions on SO concerning backing up EC2, and it has provided me with a good basis, but I'm still a bit confused as to how I should approach my solution and develop a contingency plan. Most questions are fairly specific, but I have a pretty vanilla setup, and I think this information will be beneficial to future users. Let me provide my basic setup:

  • Basic small instance
  • Pushing files to S3
  • Running MongoDB
  • Running nginx

Now, due to the ephemeral nature of EC2, it's apparent that I need to bind my EC2 instance to EBS to ensure persistent storage. The reason I'm attempting to develop a contingency plan is that I'm worried my instance may disappear at any time (due to outages, etc). If my instance were to disappear, I'm concerned that I would have to spin up a new instance and reinstall all my applications before getting everything up an running again. A few questions:

  • How do I backup my instance to ensure that if it were to disappear, I could quickly bring it back up (preferably without having to reinstall all previous software)? I don't need a series of backups, just the previous days (or weeks) backup to ensure that a previous working version exists that can be started quickly.
  • If I use EBS instead of instance storage, it essentially acts in place of my hard drive right? So if I have MongoDB installed, I'm assuming the database it is writing to cab placed on EBS?
  • If I go with the small instance with 160 GB of storage and I use EBS, would I need to allocate 160 GB of EBS out of the gate, or is the 160 GB for instance storage only?

So, in summary, I would like a solution (either manual or automatic) that can create a snapshot of my EC2 instance to ensure that if it were to disappear, it could be reconstructed without having to spend the time to manually reconstruct everything.

In an ideal world, if my instance were to disappear, I could spin up an version of my instance with everything intact (to a point where things were backed up). Any resources or suggestions? Thanks in advance.

Solution

  • OK, here we go:

    For backups:

    1. Create your instance from one of the stock AWS images. Make sure it is an EBS-backed VM - depending on the size of the VM you pick, you'll get a volume assigned with 'n' GB of space, attached as the boot volume (/dev/sda1).

    2. Configure the VM with whatever software you need, apply patches, tune for disk fragmentation, CPU consumption (task priorities, etc), and any other configuration you need that makes the VM tailored to your requirements.

    3. Stop the VM and take a snapshot of the EBS volume, then restart it (re-assigning the Elastic IP is there is one). This is your backup snapshot - repeat as desired at whatever frequency you like. Remember to stop the VM when you snap it to prevent the OS writing to the volume whilst you're taking a copy of it.

    For recovery:

    1. Your VM will fail, eventually. You'll break something and render it damaged or inoperative, or the hardware it's running on will suffer a fault. It will happen.

    2. When it does, terminate it (if it didn't self-terminate) and spin-up a new VM of the same type from the AWS stock list. Wait until it shows as 'Running', and then stop it.

    3. Detach its EBS volume and delete it.

    4. Create a new EBS volume from whichever backup snapshot you last created, and attach that new volume to the VM as /dev/sda1.

    5. Start the VM and assign your EIP if appropriate.

    About EBS storage:

    1. It's a chunk of storage space. If you format it to look like a standard disk, you can use it exactly as you would a physical disk. Install stuff on it, point software at it for use as storage space, whatever.