# Take a snapshot before risky changes: creating and restoring restore points

https://www.hostimor.com/en/knowledgebase/article/snapshots-restore-points-before-risky-changes

A snapshot leaves you a point you can return to within minutes before a risky job such as a release upgrade or a control panel installation. This guide explains how VMware snapshots work, a safe workflow, what a restore erases and why a snapshot does not replace a backup.

## How does a VMware snapshot work?

The moment you take a snapshot, the server's virtual disk is frozen and becomes read-only. Every write from then on goes to a separate file called a delta disk; reads check the delta disk first and fall back to the base disk if the data is not there. When you restore, the delta disk is discarded and the server returns to the frozen state. When you delete the snapshot, the changes in the delta disk are written into the base disk; this process is called consolidation.

This design has two consequences. First, the delta disk grows with every write; on a write-heavy server it can take up gigabytes within a few days. Second, every disk operation passes through an extra layer, so performance drops, and the larger the delta disk grows, the longer deletion takes. A snapshot is therefore a short-term safety net, not a permanent archive.

A snapshot is crash-consistent: the captured disk looks like the disk of a server that lost power at exactly that moment. File systems such as ext4 and XFS and journaling databases such as InnoDB recover from that state, but data that was still in memory and not yet written to disk is not in the snapshot.

## The Snapshots tool in the panel

With the **Snapshots** tool in the client area you can create, restore and delete snapshots. You can keep at most 2 snapshots at a time; if the limit is reached, delete an older one before taking a new one.

The panel names each snapshot itself, after the date and time it was taken, for example `snap-20261121-143005`. Write down why you took it (for example, before the PHP 8.4 upgrade) together with that date and time in your own change notes; two weeks later you will not have to guess which point you are about to return to.

## When should you take a snapshot?

A snapshot is most useful before changes that are hard to undo and whose outcome you can see right away:

- An operating system release upgrade, for example moving from Ubuntu 24.04 to 26.04 with `do-release-upgrade`.
- Installing a control panel such as cPanel, Plesk or DirectAdmin ([control panel guide](https://www.hostimor.com/en/knowledgebase/article/control-panel-on-your-server-cpanel-plesk)).
- A major version change of PHP, the database or the web server.
- A kernel update, especially if you use custom kernel modules ([updates guide](https://www.hostimor.com/en/knowledgebase/article/automatic-security-updates-on-linux)).
- Extensive changes to the firewall, SSH or network configuration.

There are also cases where a snapshot is the wrong tool: using it instead of daily backups, keeping it for weeks and taking one right before a disk expansion. Snapshots must be deleted before a disk expansion or a server migration.

## A safe workflow, step by step

1. **Make room:** Check how many snapshots the Snapshots tool shows; if there are two, delete the one you no longer need.
2. **Prepare for consistency:** On servers running a database, the cleanest approach is to stop the database briefly while the snapshot is taken. If that is not possible, pick the moment with the least traffic and take a database dump right before; that way the snapshot also contains a consistent dump.
3. **Take the snapshot:** Create it with a meaningful name and wait until it has finished.
4. **Make the change:** Apply the upgrade or installation.
5. **Verify:** Check the services, the logs and the application.
6. **Decide:** If something is wrong, restore; if all is well, delete the snapshot within a few days.

Two options for step 2:

```
# Option 1: a consistent dump right before the snapshot (database keeps running)
mkdir -p /root/dumps
mariadb-dump --single-transaction --routines --triggers --all-databases \
  | gzip > /root/dumps/before-$(date +%F-%H%M).sql.gz

# Option 2: stop the database while the snapshot is taken
systemctl stop mariadb
sync
# ... take the snapshot in the panel; when it has finished:
systemctl start mariadb
```

Basic checks you can use in step 5:

```
# Failed services and errors in the current boot
systemctl --failed
journalctl -p err -b --no-pager | tail -n 30
# Does the site respond?
curl -sI https://example.com | head -n 1
```

## What to know before you restore

A restore permanently erases everything written to disk after the snapshot. That covers not only the broken configuration but also orders, user registrations, uploaded files, received email and logs from the time in between. On a live site, export the data created after the snapshot to a location outside the server before you restore:

```
# Before restoring: export the data that arrived after the snapshot
mariadb-dump --single-transaction shop orders customers | gzip > /root/arrived-later.sql.gz
# Run on your own computer or a backup server
rsync -avz root@203.0.113.10:/var/www/example.com/uploads/ ./uploads-copy/
scp root@203.0.113.10:/root/arrived-later.sql.gz ./
```

Replace the database and table names in the example with your application's. After the restore, re-import these records if needed, and watch for primary key conflicts so the same records are not added twice.

A restore brings back the disk as it was at the moment of the snapshot, while the server's CPU and RAM are set again to your current plan. If the server was running before the restore, the panel starts it again and it boots with the disk as it was at the moment of the snapshot. Application sessions started and access tokens issued after the snapshot may become invalid, an SSL certificate renewed in the meantime reverts to the old one, and security updates applied in between are rolled back. After the restore, check the clock, the services and the updates:

```
timedatectl            # should show System clock synchronized: yes
systemctl --failed
# Reapply updates released after the snapshot
apt update && apt upgrade -y      # AlmaLinux: dnf upgrade -y
```

> ****Warning****
> 
> A restore cannot be undone. Of the data written after the snapshot, only what you exported can be recovered; if in doubt, take the dump first and restore afterward.

## Limits and performance

- **Number:** You can keep at most 2 snapshots at a time; when the limit is reached, delete an old one to take a new one.
- **Disk expansion and migration:** All snapshots must be deleted before a disk expansion (a package upgrade or an additional disk) and before a server migration; a disk cannot be expanded while snapshots exist. See the [resource upgrade](https://www.hostimor.com/en/knowledgebase/article/resource-upgrade-cpu-ram-disk) guide for details.
- **Duration:** The longer a snapshot exists, the larger the delta disk grows, the slower the disk gets and the longer deletion takes. Do not keep one for more than a few days.
- **When to delete:** During deletion the changes are written into the base disk, so the disk may be briefly slower; delete outside peak hours.

## A snapshot is not a backup

A snapshot sits on the same storage as the server itself. If the server is deleted, for example after it enters the deletion queue following an unpaid invoice ([overdue invoice timeline](https://www.hostimor.com/en/knowledgebase/article/what-happens-if-a-server-invoice-is-overdue)), its snapshots go with it, and a problem in the storage layer affects both at once. A snapshot also only brings back the whole disk: to recover a single file or a single database, you have to restore the entire server.

Unless your package explicitly includes backups, backing up your data is your responsibility. For an encrypted, versioned backup outside the server, follow the [backup strategy with restic](https://www.hostimor.com/en/knowledgebase/article/server-backup-strategy-restic) guide, and read the [Backup and Data Retention Policy](https://www.hostimor.com/en/backup-and-data-retention-policy) for the scope.

|  | Snapshot | External backup (restic, etc.) | Reinstall |
| --- | --- | --- | --- |
| What does it protect? | The whole disk at a single moment | The files and database dumps you choose, in many versions | Nothing; it gives you a clean system |
| Where is it stored? | On the same storage as the server | Outside the server, in another location | — |
| If the server is deleted | It is lost | It survives | — |
| Recovery | Within minutes, the whole server | Selectively, per file or database | Within minutes; all data is erased |
| When to use it | Right before a risky change, for a few days | Continuously, automated and regularly tested | When the system is broken or you change releases |

For details on reinstalling, see [reinstalling the operating system](https://www.hostimor.com/en/knowledgebase/article/reinstalling-the-operating-system-guide).

## Frequently asked questions

### Does the server stop while a snapshot is taken?

The server keeps running. Disk operations may feel briefly slower while a snapshot is created and especially while it is deleted, so choose a time outside peak hours.

### I cannot take a third snapshot.

You can keep at most 2 snapshots at a time. Delete the snapshot you no longer need, then take the new one.

### Why does a disk upgrade require me to delete my snapshots?

VMware does not allow a virtual disk to be resized while delta disks are attached to it. Delete the snapshots before the upgrade; once the expansion is complete, take a new snapshot if you need one.

## Related articles

- [A server backup strategy: the 3-2-1 rule, off-site backups with restic and database dumps](https://www.hostimor.com/en/knowledgebase/article/server-backup-strategy-restic)
- [What changes on your server after a CPU, RAM or disk upgrade?](https://www.hostimor.com/en/knowledgebase/article/resource-upgrade-cpu-ram-disk)
- [Reinstalling the operating system: which OS to choose and what to do first](https://www.hostimor.com/en/knowledgebase/article/reinstalling-the-operating-system-guide)
- [Automatic security updates and rebooting after a kernel update](https://www.hostimor.com/en/knowledgebase/article/automatic-security-updates-on-linux)
- [Your first 30 minutes on a new VDS or VPS: a post-setup checklist](https://www.hostimor.com/en/knowledgebase/article/first-30-minutes-on-a-new-vds-or-vps-checklist)
