Shrinking a Root EBS Volume

This guide explains how to shrink a root EBS volume on an AWS EC2

Background: We have an EC2 instance with a root EBS volume that we want to shrink

Step 0: We'll call the existing EC2 instance IA and the existing EBS volume VA

Step 1: Create a new EC2 instance (We'll call this IB) using the AMI from IA (You can do this by clicking the Instance details, click the AMI ID, and clicking Launch instance from AMI) and the same type (e.g. r6gd.2xlarge). When selecting the root drive for IB, select the same type as VA and make it the target size that you'd like to shrink VA to, we'll call this drive VB

Step 2: SSH into IB and upgrade/downgrade the OS to match IA as needed. Ensure the welcome messages match on both IA and IB. (e.g. Welcome to Ubuntu 20.04.5 LTS (GNU/Linux 5.15.0-1019-aws x86_64))

Step 3: Stop IA and IB from the EC2 Console

Step 4: Detach VB and VA from their instances

Step 5: Create a snapshot of VA

Step 6: Create a new volume of the same type and size of VA from the snapshot, we'll call this volume VA2

Step 7: Create another new EC2 instance using the same method as Step 1, we'll call this instance IC. (The root volume for IC can be a small 8gb basic gp3 drive, we're just using this to copy between VA2 and VB)

Step 8: Attach the volumes to IC in the following order:

• Attach VB as /dev/sdf

• Attach VA2 as /dev/sdg

Step 9: Create the directories using sudo mkdir /source /target

Step 10: Check the drive names of VB and VA2 with sudo lsblk

• In my case it was /dev/nvme2n1p1 for VB and /dev/nvme3n1p1 for VA2

Step 11: Format VB using sudo mkfs.ext4 /dev/nvme2n1p1

Step 12: Mount VB to target using sudo mount -t ext4 /dev/nvme2n1p1 /target

Step 13: Mount VA2 to source using sudo mount -t ext4 /dev/nvme3n1p1 /source

Step 14 [VERY IMPORTANT]: We need to properly label the file system with an e2label in order for Linux to boot correctly

• First, check the label of VA2 using sudo e2label /dev/nvme3n1p1

• In my case it returns cloudimg-rootfs

• Next, we'll set VB to have the same label using sudo e2label /dev/nvme2n1p1 cloudimg-rootfs

Step 15: Now we can copy the contents using sudo rsync -vaxSHAX /source/ /target Note: there is no "/" following "/target"

Step 16: When the files are finished copying, unmount VB using sudo umount /target

Step 17: Detach VB from IC

Step 18: Attach VB to IA as /dev/sda1

Step 19: Restart IA

Step 20: After verifying that IA is working correctly with VB. Cleanup and delete the created instances and volumes (IB, IC, VA2, VA)

-----------------------------------------------------------------------------------------------------

[Optional] Using Instance Store Swap: If you used a swapfile on IA there's a chance that it won't work correctly after changing volumes from VA to VB. To fix this we can follow these steps:

Background: My swapfile is mounted on /eph located on the instance store volume at /dev/nvme1n1

Step 1: Delete the existing swapfile on /eph

Step 2: Calculate the number of MB you want to use for your swapfile, in my case I used 34360MB

Step 3: Run the following commands:

sudo dd if=/dev/zero of=/eph/swapfile bs=1M count=34360
sudo chown root:root /eph/swapfile
sudo chmod 600 /eph/swapfile
sudo mkswap /eph/swapfile
sudo swapon /eph/swapfile

Step 4: Ensure fstab is set correctly sudo nano /etc/fstab

• If there isn't an entry for the swapfile add the following: /eph/swapfile swap swap defaults 0 0

Step 5: Ensure the swapfile initializes at start with sudo swapon -a

Step 6: Reboot the instance to ensure that the swapfile initializes correctly