Technical Anshu
Tutorials Blog
Snapshots

Take an ESXi Snapshot Before Upgrading Ubuntu and Never Keep It Forever

A snapshot before a big Ubuntu upgrade is one of virtualization's genuine superpowers: if the upgrade goes wrong, you rewind the whole machine. But the same feature, left running for months, quietly eats your datastore and slows the VM it was protecting. This page teaches the ritual - and the discipline that has to come with it.

By Technical AnshuVMware ESXiSnapshots done rightNot a backup
Watch the full video - Ubuntu virtualization on ESXi, the setup, configuration and troubleshooting, step by step

Here is the upgrade scene every administrator knows: the update runs, something deep breaks, and the evening disappears. An ESXi snapshot taken five minutes earlier turns that story into a shrug - revert, and the machine is exactly as it was. The catch is that a snapshot is a living thing: from the moment you take it, the VM's changes pile up in a side file that grows until you deal with it. Snapshot literacy is two skills - taking them at the right moment, and closing them on purpose.

Snapshot Manager - a healthy picture
NameBefore Ubuntu upgrade - with the date
TakenJust before the risky change
Memory state includedUsually no - cleaner and lighter
Planned lifeDays, while the change proves itself
Kept for monthsGrows, slows the VM, risks the datastore
Example snapshot view - one fresh safety net, not a museum of old ones
1

What a Snapshot Actually Is

A snapshot freezes the VM's disk state at a moment. From then on, the VM writes its changes into a new delta file, and the original disk sits untouched underneath. Revert throws the delta away and you are back at that moment; delete (consolidate) folds the delta into the base disk and you keep the present.

Two consequences flow from that mechanism. First, while a snapshot lives, the VM's storage work splits across files - long snapshot chains cost real performance. Second, the delta grows with every change the VM makes, at the VM's own pace: a busy server can grow its snapshot by gigabytes a day. The snapshot is a tab being run up against your datastore, and the bill arrives when you consolidate - or when the datastore fills first.

2

The Before-Upgrade Ritual

  1. Make sure open-vm-tools is installed in the Ubuntu VM, so the snapshot can be quiesced - the guest settles its filesystems instead of being frozen mid-write
  2. For an OS upgrade, the cleanest snapshot is of a powered-off VM. If uptime matters, snapshot while running but leave "include memory" unchecked - a disk-only snapshot is smaller and is what an upgrade rollback actually needs
  3. Name it like you mean it: what change, and the date. "Before Ubuntu 24.04 upgrade - Oct 2026" will still make sense in a tired moment; "test1" will not
  4. Run the upgrade: sudo apt update and sudo apt upgrade for routine updates, or the release-upgrade path for a version jump
  5. Test the machine's real job - the service, the shares, the logins - not just "it boots"
3

Close It on Purpose - Delete or Revert

Once the upgraded VM has proved itself - a few days of normal work is a fair trial - close the snapshot deliberately:

  1. Delete the snapshot to keep the upgraded state. ESXi consolidates the delta into the base disk; on a busy VM this takes time and generates storage load, so do it in a quiet window, not during the office's busiest hour
  2. Revert only when the change failed. Everything since the snapshot - including any data the VM wrote - is discarded. That total amnesia is the feature and the danger in one

If you find yourself afraid to delete a snapshot "just in case", hear what that fear is telling you: the VM needs a real backup. Which is the next section.

4

A Snapshot Is Not a Backup - Say It Plainly

The snapshot lives on the same datastore, in the same host, as the VM it protects. A dead datastore, a dead host or a ransomware event that reaches the hypervisor takes the VM and its snapshots together. A backup is a copy somewhere else.

So the standing pair of rules:

  1. Snapshots are for changes - upgrades, experiments, risky configuration work. Lifespan: days
  2. Backups are for disasters - an exported copy or a backup job writing to separate storage. The OVF guide in this series shows the simple standalone-host version

Keeping a snapshot "as the backup" for six months delivers the worst of both: no disaster protection, and a datastore slowly filling with delta files.

5

The Slow Damage of Old Snapshots

If you inherit a host, audit snapshots early. The symptoms of snapshot debt: a datastore shrinking although the VMs look small (their delta files are not small), a VM that feels slower each month (its disk chain is long), and consolidation jobs that run for hours because months of changes must be merged.

The recovery is patience, not heroics: pick a quiet window, delete the oldest snapshots one at a time, let each consolidation finish, and watch free space return. Then set the household rule that prevents relapse - every snapshot gets a name, a reason and a planned deletion date, and the calendar is honoured.

🔴 The two-line rule: snapshot before the change, delete after the proof. A snapshot older than a couple of weeks is not a safety net anymore - it is a liability with a progress bar.

Watch Ubuntu on ESXi in the Video

The full process on a live ESXi host - creating the virtual machine, installing Ubuntu and fixing the common problems, exactly as they appear on screen.

Watch Full Video
Frequently Asked Questions

How long can I safely keep an ESXi snapshot?

Think in days, not months - long enough for the change it protects to prove itself, commonly a few days to about two weeks for an upgrade. The snapshot's delta file grows with every write the VM makes the whole time it exists, costing datastore space and disk performance. Once you trust the upgraded system, delete the snapshot and let it consolidate.

If the host or datastore dies, will the snapshot save my VM?

No. Snapshots live on the same datastore as the VM itself, so the same failure takes both. A snapshot protects against a bad change - a failed upgrade, a broken setting. Disaster protection needs a copy on separate storage: a backup job or an exported copy of the VM, as the OVF move guide in this series describes.

What is the difference between deleting a snapshot and reverting to it?

Deleting keeps the present: the changes recorded since the snapshot are merged into the VM's main disk and the snapshot disappears. Reverting discards the present: the VM returns to exactly the state captured in the snapshot, and everything written since - files, settings, data - is gone. Delete is the happy ending; revert is the ejector seat.

This guide is based on the Technical Anshu video on Ubuntu virtualization on ESXi - stuck on a step? Ask in the video comments and you will get a reply.