Why?There was a reason for everything 🛠️

Hello everyone! Today I’d like to share my recent experience with VM disk resizing issues in my MoolMeow project’s Kubernetes setup. This is a common challenge many developers face when working with Vagrant and VirtualBox! 🤔

📌 The Problem

I was building a Kubernetes cluster on my Mac Mini using Vagrant and VirtualBox. I wanted to allocate 128GB of disk space to the master node and 64GB to the worker node.

# Vagrantfile configuration
NODES = [
  { hostname: "k8s-master", ip: "192.168.56.10", memory: 8192, cpus: 4, disk_size: "128GB" },
  { hostname: "k8s-worker", ip: "192.168.56.11", memory: 4096, cpus: 2, disk_size: "64GB" }
]

I installed the vagrant-disksize plugin and provisioned the VMs, but when I checked the disk space inside the VM, it was still at the default size (around 60GB)! 🧐

$ df -h /
Filesystem                         Size  Used Avail Use% Mounted on
/dev/mapper/ubuntu--vg-ubuntu--lv   60G  5.2G   51G  10% /

What went wrong? 🤔

🧩 The Root Cause

There were two main issues:

1. Disk Expansion Is Not Partition Expansion

When you set the disk_size in Vagrant, it only increases the physical disk size, but partitions and filesystems don’t automatically expand. Ubuntu VMs use LVM (Logical Volume Manager), creating a layered structure:

Physical disk(/dev/sda) → Partition(/dev/sda3) → PV → VG → LV → Filesystem(ext4)

I needed to expand each of these layers sequentially!

2. Ansible Automation Challenges

When trying to automate this process with Ansible, the parted command didn’t work properly because:

• parted displays a Fix/Ignore? prompt when modifying GPT disks
• Handling interactive prompts in Ansible is tricky
• The command and response input sequences got confused, resulting in an "invalid token" error

🛠️ My Troubleshooting Journey

After several attempts, I found a solution!

Failed Attempts ❌

1. Using parted with the --pretend-input-tty flag and heredoc:

   parted /dev/sda --pretend-input-tty <<EOF
   Fix
   resizepart 3
   100%
   Yes
   quit
   EOF
   

2. Passing inputs with echo -e:

   echo -e "Fix\nresizepart 3\n100%\nYes" | parted /dev/sda --pretend-input-tty
   

These approaches failed because of input sequence issues. 😞

The Working Solution ✅

The key insight was to separate the GPT table modification from the partition resizing:

# master.yml (Ansible playbook)
- name: Fix GPT table first
  shell: |
    echo "Fix" | parted /dev/sda ---pretend-input-tty print
  register: gpt_fix
  ignore_errors: true
  changed_when: false

- name: Resize partition 3 to use all available space
  shell: |
    echo "100%" | parted /dev/sda ---pretend-input-tty resizepart 3
  register: parted_result
  changed_when: "'Error' not in parted_result.stdout"

Then I expanded each LVM layer sequentially:

- name: Resize Physical Volume
  shell: pvresize /dev/sda3

- name: Extend Logical Volume
  shell: lvextend -l +100%FREE /dev/mapper/ubuntu--vg-ubuntu--lv
  register: lv_result
  changed_when: "'extended' in lv_result.stdout"

- name: Resize filesystem (ext4 specific)
  shell: resize2fs /dev/mapper/ubuntu--vg-ubuntu--lv
  register: fs_result
  changed_when: "'resized' in fs_result.stdout or 'nothing to do' in fs_result.stdout"

🎉 The Results

Success at last! Checking the disk space:

$ df -h /
Filesystem                         Size  Used Avail Use% Mounted on
/dev/mapper/ubuntu--vg-ubuntu--lv  123G  5.2G  112G   5% /

The master node was now at 128GB and the worker node at 64GB, exactly as intended! 🙌

📝 Key Lessons Learned

This troubleshooting process taught me several valuable lessons:

1. Virtual Disk Expansion ≠ Filesystem Expansion: Increasing a VM’s disk size is a multi-step process.
2. Interactive Command Automation Challenges: Commands like parted require special handling when automated.
3. LVM Understanding is Critical: Understanding Linux volume management architecture makes troubleshooting much easier.
4. The Power of Command Separation: Breaking down complex interactive commands into simpler sequential steps is effective.

🔍 Complete Resizing Workflow Summary

The entire process can be summarized as:

1. Disk Expansion: Configure disk_size in Vagrant
2. GPT Table Modification: Use the Fix option in parted
3. Partition Expansion: Expand /dev/sda3 with parted resizepart
4. PV Extension: Run pvresize /dev/sda3
5. LV Extension: Run lvextend -l +100%FREE /dev/mapper/ubuntu--vg-ubuntu--lv
6. Filesystem Extension: Run resize2fs /dev/mapper/ubuntu--vg-ubuntu--lv

Now with this automated in Ansible, I can provision VMs with the desired disk sizes every time! 😄

If you’re facing similar issues with Vagrant or other virtualization environments, I hope this post helps. Feel free to share your own tips or questions in the comments! 💬