Security

You have a backup. Can your business get back to work?

There is a green tick next to the latest backup. The file was created today, and your data appears to be safe. You still need to establish whether that backup can restore your website or business system well enough for the company to resume work.

Author
CreateIT
Published
Reading time
12 min

Listen to the articlesynthesised voice, 15:52

Start reading
Image 1Fictional example. Eleven hours pass between the 03:00 backup and the 14:00 outage. Recovering data from that interval requires other sources.

First, check that the backup was created. Then establish whether you can access it, whether it contains the data you need and whether you can use it to restore the system. Check that your team can also carry out its work in the restored system. Even an undamaged backup may be too old if recent records are missing.

A test may reveal that only one person knows the password, or that the encryption key needed to read the data is missing. You may restore the database successfully, then discover that files uploaded by customers were never backed up. These problems are easier to address during a planned test while your everyday system is still running.

What happened to the orders placed after the last backup?

Imagine an online shop whose full backup was created at 03:00. At 14:00, the server becomes unusable. This is a fictional example, not a CreateIT client case.

The team restores the overnight backup, and the shop's homepage loads again. But that backup contains no orders received after 03:00. It reflects the data as it stood 11 hours before the outage. Without another verified recovery source, that backup alone cannot recover the newer records.

Payment provider records may help recover information about some completed payments. The warehouse system may still hold information about dispatched orders. However, those records need to be compared and restored in the right sequence. A payment alone cannot tell you whether the warehouse has already shipped an order.

A working homepage is therefore not enough to consider recovery complete. You need to know the point in time to which data has been recovered, which records remain missing and which actions must not accidentally be carried out twice.

Two questions to answer before an outage

First, agree how quickly the required service must be restored after an interruption. This is the recovery time objective (RTO). Allow time for gaining access, preparing the environment and checking that the service works again, as well as loading the database onto the server.

The second question concerns data. How much time's worth of data can the business afford to lose? The recovery point objective (RPO) helps establish how recent the recoverable data must be. If the business can tolerate only a small gap, one overnight backup may be insufficient.

RTO and RPO are targets agreed in advance. Compare them with the results of a test. If you agreed to restore the service within four hours but the exercise took eight, investigate and address the cause of the delay.

The definitions of RTO and RPO and the principles of recovery planning are explained in NIST's contingency planning guidance. This 2010 document was written for US federal information systems; this article draws on its planning concepts.

What needs to be restored alongside the database?

A website needs more than a database containing customer details, orders and content. Images, documents, application code, software libraries and configuration may be stored separately. You also need access to external services and someone who knows how to start the system.

WordPress documentation describes database and file backups separately. Exporting a database to a file does not create a backup of the entire website. Copying the website's directory does not, by itself, preserve the database either.

Both parts are explained in the WordPress backup documentation.

The restored components also need to be consistent with one another. If the database refers to a document uploaded moments ago but the file backup was made earlier, that document may be missing. Choose a backup method that fits the way the system operates, its database type and the sequence of changes.

For example, the PostgreSQL documentation on file-level backups explains why you cannot simply copy the files of a running database without an appropriate method. Consult the documentation for the database you actually use.

This does not mean putting every password in an archive alongside the website. Store credentials and encryption keys securely. The recovery plan should explain how an authorised person can obtain them even when the main system is no longer working.

Will the backup remain accessible if you lose access to your account?

A backup in another directory on the same server may help restore an accidentally changed file. It can also disappear with the server. Similarly, backups in a different storage location may still be at risk if one compromised account can access and delete both the working system and its backups.

CISA recommends keeping offline, encrypted backups of critical data and regularly checking their availability and integrity. An offline backup is not continuously reachable from the working system. This reduces the chance that an attack on that system will also destroy every recovery source.

This approach is described in the CISA StopRansomware Guide.

Another protective measure is a backup that cannot be overwritten or deleted during a defined retention period. Services often call this an immutable backup. Check the rules of the particular mode, administrators' powers and the protection of encryption keys. Losing a key can make the stored encrypted file unusable.

Find out who can change the retention period and who receives a notification when it changes. A product name or an enabled setting is insufficient. You need to understand who can access the backups and how they will be used for recovery.

Protection of backups, access permissions and encryption keys is covered by the NCSC principles for ransomware-resistant on-premises backups. The available controls depend on the chosen solution.

Start the exercise in a separate environment

Do not overwrite the live website during a test. Prepare a separate environment with restricted access and restore the backup there. Decide in advance which connections may be used and which must be blocked.

The restored copy may retain scheduled tasks and settings that allow it to send email and connect to your payment, warehouse or accounting systems. Isolate the test environment before starting the application. Block connections to production systems or replace them with connections to services intended for testing. The test environment must not send old emails to customers, charge payments or submit orders for fulfilment again.

Use email addresses controlled by the people carrying out the test and suitable test services. If the backup contains personal data, restrict access to the test environment and define how that data will be deleted. Preventing search engine indexing does not, by itself, protect access to the data.

Record what the exercise will test. Using a payment provider's test environment allows you to check the connection logic. It does not establish that the production settings are correct. The report must also identify what remains untested.

The AWS recovery testing example likewise distinguishes completion of a restore job from validation of the result, and recommends a test or isolated environment for the particular database exercise.

A restored homepage does not mean work has resumed

Before the test, choose the activities your business needs in order to keep operating. For an online shop, that could mean finding and processing an order. For a service business, it might mean receiving an enquiry. In an employee portal, check that a user can sign in and open the documents they need.

Examples of recovery checks and supporting evidence
What to checkUseful evidence
Data recovery pointA specific time is recorded. Selected records before and after that point have been checked to establish which were restored and which are missing.
Files and linked recordsSelected documents and images referenced by the system can be opened.
AccessAn authorised user can sign in to their account and access only the data they are permitted to use.
Customer enquiryA test enquiry reaches the intended record-keeping system and the responsible employee.
Connections to other systemsThe test environment receives the expected response without affecting real customer orders or payments.
Resuming workA colleague familiar with the process tries their work task and identifies anything still unclear or not working.

A checksum is a value calculated from a file's contents. Comparing it with a trusted, previously calculated checksum can help establish whether the contents have changed. Matching checksums do not establish that the archive contains every component the business needs or that the application will work. Both the data and the restored system need to be checked.

If time and resources do not allow a full exercise, start by testing a limited set of activities. State exactly which activities were tested and what was left out. Recovering one file does not establish that the entire service could be restored.

After an attack, choosing yesterday's date is not enough

If the system stopped following a compromise, a backup may contain changes made by the attacker. The attack may have begun before the first visible fault. Choose a backup based on the investigation's findings and check its contents before restoring it.

Address the weakness that allowed the attacker into the system and establish which passwords, keys and accounts may have been exposed. Otherwise, the same problem could threaten the restored system. Coordinate evidence preservation, containment and recovery with the people investigating the incident.

Checking recovery sources, addressing the causes of the incident and verifying restored resources are included in NIST incident response guidance SP 800-61r3.

If you suspect an active attack, do not rush to make a website restored from an old backup publicly accessible. First, establish which systems are affected and agree the response with the incident investigators.

What to record after a recovery exercise

Document the work so that someone else can repeat it. Identify the backup used, the point in time its data represents, the restored system version, the steps performed and their results. Do not put passwords or encryption keys in the report.

You can complete the attached worksheet during the exercise. It is a blank template and contains no results from an actual recovery test. Download the recovery test worksheet.

Agree in advance when you will start measuring recovery time and what will count as completion. Record separately the time spent waiting for access, obtaining data, restoring the environment and performing checks. This shows where the technical solution needs improvement and where responsibilities need to be clearer.

The time measured in a test describes that environment and those circumstances. Conditions during a real incident may differ. The required person may be unavailable, or the service or infrastructure used in the test may be inaccessible. A test result is therefore not a guarantee that the system can always be restored just as quickly.

Agree who will address each problem and by when. After the fix, test the affected activity again. Set the frequency of exercises according to the system's importance, changes to it and the acceptable amount of data loss. A significant change in storage or software is a reason to review the recovery procedure too.

Five questions for your next conversation with your maintenance provider

  1. Which business activity can we resume by restoring this backup? Ask which system components are required and what the backup excludes.
  2. To what point in time can we recover the data? Find out how to handle later records and where other recovery sources may be available.
  3. How will we obtain the backup if the main account is unavailable? Establish who can access the backup and encryption keys, and who can stand in for the responsible colleague.
  4. When did we last try this activity after a restore? Ask for the test record showing its scope and results.
  5. What still prevents us from meeting the business's recovery target? Agree specific improvements and a repeat test.

You can ask these questions without being familiar with the underlying technology. Begin by establishing what the current solution can restore and what remains unknown. That gives you a basis for deciding whether you need a different service.

We can take over and maintain your website

CreateIT also takes over websites built by other developers. We examine the code, versions, server, database and integrations, and organise access permissions and backups. We work with WordPress, WooCommerce, Drupal and custom systems.

We can begin by discussing which activity your business needs to resume after a system outage and agreeing an appropriate test. We define the maintenance scope, responsibilities and required work after examining the system. Recovery options and timing depend on the system's condition and the available data.

In your first message, include the website address, the system it uses and what you know about backups and previous tests. Do not send backups, passwords or encryption keys with the initial enquiry. If we need them for the work, we will agree a suitable access procedure.

Read more about working with us in our website support and maintenance service description.

Share

LinkedIn Facebook Email

Newsletter

New articles in your inbox

Sign up to the CreateIT newsletter for practical articles on development, AI, marketing and cybersecurity.

We plan to send the newsletter no more than once a month.