Fear not, we are here to clear things out.

In order to understand where the problem is, we first need to make a clear distinction between two different kinds of backups: inconsistent backup and consistent backup.

Let’s go deeper into this two key-concepts.

Inconsistent Backups: VM Snapshots

When talking about Virtual Machines (from now on VMs), a Snapshot is a point-in-time photograph taken at the Hypervisor level and is the composition of the state and the data of VM.

Inconsistent Backups: Crash-consistent Snapshots

So, Cloud Providers run instances on VMs, then snapshots taken should be VM Snapshots, right? Wrong.

Snapshots created by Cloud Providers, like AWS, with API calls or from the web console are taken instead at the block-storage level, so they are basically photographs of the file-system at that moment. This kind of snapshots are much less complex than VMs Snapshots and comes with many features like less space occupied (as the state is not taken into account), the ability to get only the delta of the data that changed, and the ease of management.

The downside is that the state is not saved and, consequently, it’s not possible to recreate the exact state of the machine at the time the snapshot was taken.

The main feature is that all of the data on the VM is saved at the same time, so this should let us sleep peacefully, right? Wrong again.

This kind of Snapshots leverage the current file-systems ability (i.e., Ext4, NTFS) to journal file changes and recover from inconsistent data without leaving them corrupted while Databases use transaction logs to return to the last consistent state even in the Snapshot was taken in the middle of a transaction.

Most of the time everything will work out by itself, but sometimes you find yourself again yelling at the Cloud.

So how do you prevent this kind of errors?

Application-Consistent Backups

The one and the simplest answer is that something needs to be done before a backup to ensure that all data is consistent. The most common way to do this is to stop all I/O operations for the file-system and to prevent writing to the Databases. This comes at a price though; the system needs to run some routines and scripts before the snapshots could be taken therefore the application will be in a read-only state for a brief period. Other than that, now you need a way to start scripts on the VM and maintain them.

In conclusion…

you’re not bound to go through the trouble of setting up consistent snapshots of your infrastructure and, in the end, it’s up to you to decide what kind of backup you need.

At least now you have the tools to understand what could go wrong and stop yelling at Cloud.

If you have concerns on your backup strategy, Smart Backup is here to help!


Nicolò Marchesi
DevOps Engineer at Noovolari Smart Backup

Lascia un commento