On this page

A disaster recovery plan template gives you the sections every recovery plan needs, so you fill in your own systems instead of starting from a blank page. The finished plan answers four questions: what must be restored, how fast, how much data you can afford to lose, and who does what while it happens.

This guide gives you a template to copy, a worksheet for the two numbers that drive it, and a worked example. If the worksheet shows that your backups or architecture cannot meet those numbers, Cloud consulting services is where we review the architecture with you.

What a disaster recovery plan template covers

A recovery plan template covers five things: the systems in scope, a recovery target for each, where the backups are, the steps to restore each system, and a record of tests. Everything else in the plan supports those five.

The reference most templates descend from is NIST Special Publication 800-34 Revision 1, the Contingency Planning Guide for Federal Information Systems, from the US National Institute of Standards and Technology (NIST). It was published in 2010, and NIST still lists it as the final version as of November 2026. It defines a disaster recovery plan as an information system-focused plan designed to restore a system, application or computer facility at an alternate site after an emergency.

NIST organizes a plan into three phases, and the template below follows them:

  1. Activation and notification. Who declares a disaster, on what evidence, and who is told.
  2. Recovery. The ordered steps that bring each system back.
  3. Reconstitution. Testing the restored system, returning to normal operation and closing the event.

How it differs from an incident response plan

An incident response plan covers detecting, containing and communicating about a security incident. A disaster recovery plan covers restoring systems and data after a major outage, whatever caused it. The two meet at recovery, so each should point to the other. Our incident response plan template covers the first of the two.

RTO and RPO: the two numbers that drive the plan

The recovery time objective (RTO) is how long a system can be down, and the recovery point objective (RPO) is how much recent data you can lose. Amazon Web Services (AWS) defines RTO as the maximum acceptable delay between the interruption of service and its restoration. It defines RPO as the maximum acceptable amount of time since the last data recovery point.

Set both per system, not once for the whole company. A payments database and an internal wiki do not deserve the same spend. NIST adds a third number, the maximum tolerable downtime (MTD): how long a business process can be disrupted without significant harm. The RTO must normally be shorter than the MTD, because work that piled up during the outage still has to be processed after the system returns.

The RTO and RPO worksheet

Fill in one row per system, starting with the question a founder can answer: what stops if this is down. The rows below are an illustrative example for a small software-as-a-service (SaaS) product, not a recommendation.

An illustrative RTO and RPO worksheet for four systems of a small SaaS product
SystemIf it is downRTORPO
Production databaseCustomers cannot use the product4 hours1 hour
File storageUploads and downloads fail4 hours24 hours
Billing jobsInvoices go out late24 hours1 hour
Internal analyticsReports are stale3 days24 hours

Two rules keep the worksheet honest. First, the people who own the revenue set the targets, and engineering prices them. Second, a target is only real once a test has met it. Until then it is a wish.

Choose a recovery strategy for each tier

The recovery strategy is how you meet the numbers, and tighter numbers cost more. The Reliability Pillar of the AWS Well-Architected Framework describes four strategies, in increasing order of cost and complexity. The bands in the table are the figures AWS publishes for its own cloud as of November 2026. They are not our commitments, and they are not yours until you test them.

The four AWS disaster recovery strategies with the RPO and RTO bands AWS publishes and what stays running in the recovery Region
StrategyRPORTOWhat stays running
Backup and restoreHours24 hours or lessBackups only
Pilot lightMinutesTens of minutesReplicated data stores
Warm standbySecondsMinutesA scaled-down full copy
Multi-Region active-activeNear zeroPotentially zeroFull copies serving traffic

For an early-stage product, backup and restore is a reasonable choice for most systems. AWS notes that automated or continuous backups with point-in-time recovery can lower the RPO to as low as 5 minutes in some cases. Keeping your infrastructure as code shortens the RTO, because you rebuild servers from a file instead of from memory.

These are AWS's names and figures, so check your own provider's documentation for its equivalents. High availability is a separate matter. Spreading servers across zones keeps you running through a hardware failure, but it will not bring back a table that someone deleted. Only a backup does that.

The disaster recovery plan template, ready to copy

Copy the template into the tool your team already uses, then replace every bracket. It follows the three NIST phases, and it works as an IT disaster recovery plan template for a product that runs in the cloud. Keep a copy somewhere that does not depend on the systems it describes.

DISASTER RECOVERY PLAN: [Company name]
Version: [n]   Owner: [name]   Last tested: [date]   Next review: [date]

1. SCOPE
Systems covered: [list]
Systems not covered, and why: [list]
Related plans: incident response plan [link], status page [link]

2. ROLES AND CONTACTS
Recovery lead (declares a disaster, makes the calls): [name, contact]
Deputy: [name, contact]
Technical owners, one per system: [system: name]
Communications owner (customers, staff): [name]
Vendor support contacts: [cloud, database, DNS, payments]

3. RECOVERY TARGETS
System | If it is down | RTO | RPO | Strategy
[system] | [impact] | [time] | [time] | [backup and restore,
  pilot light, warm standby or active-active]

4. BACKUPS
For each data store: what is backed up, how often, where the copies
live, how long they are kept, who can delete them, and the date of
the last restore test.

5. ACTIVATION AND NOTIFICATION
Declare a disaster when: [conditions, for example "primary database
  cannot be repaired in place" or "cloud region down beyond [time]"]
Who decides: [recovery lead, then deputy]
Who is told, in order: [team, leadership, customers, vendors]
Where the team meets if chat is down: [channel or phone bridge]

6. RECOVERY RUNBOOKS (one per system, in priority order)
Runbook: [system]
  Before you start: [access needed, where credentials are kept, tools]
  Steps: 1. [step]  2. [step]  3. [step]
  Verify: [checks that prove the system works and the data is intact]
  Expected time, from the last test: [time]
Never write passwords or keys in this document.

7. RECONSTITUTION
Checks before reopening to customers: [list]
Data to reconcile (entered during the outage, or lost past the RPO):
  [list]
Who declares the event over: [name]
Review meeting within: [days]

8. TEST LOG
Date | Type (restore, tabletop, failover) | System | RTO met | RPO met | Fixes

9. MAINTENANCE
Review when: a system is added or replaced, a vendor changes, a test
fails, or the scheduled review date arrives.

Section 4 is where a first draft often goes quiet. If nobody can say when a restore was last tested, write "never tested" and make it the first task.

Section 5 only works if someone notices the failure quickly. Our guide to observability vs monitoring covers what a small team needs first.

Disaster recovery plan example: a lost database

This disaster recovery plan example shows how the two numbers expose a gap before an outage does. The company and the figures are illustrative, chosen to be round.

A team runs one production database with a nightly backup at 02:00. Its worksheet says RTO 4 hours and RPO 1 hour. At 17:00 a faulty migration corrupts the main tables.

  • Data lost: the last good copy is from 02:00, so 15 hours of customer data are gone. The RPO was 1 hour. The plan had failed before anyone opened a runbook.
  • Time to restore: the last restore test took 3 hours, inside the 4-hour RTO. That half of the plan holds.

The fix is in the backup method, not the runbook. With continuous backup and point-in-time recovery, the team can restore to a moment just before the migration ran. The data lost falls from 15 hours to minutes, inside the 1-hour RPO.

The example also shows why a replica is not a backup. A standby replica would have copied the faulty migration almost at once. Recovery needs a copy from before the mistake.

Backups: the 3-2-1 rule and restore tests

A backup counts only once you have restored from it. Start with the 3-2-1 rule, as stated in Data Backup Options, a US-CERT paper hosted by the US Cybersecurity and Infrastructure Security Agency (CISA). Keep 3 copies of any important file, on 2 different media types, with 1 copy stored offsite.

For a cloud product, that becomes four plain checks:

  • The backups sit in a different account or region from the data they protect, so one mistake or one stolen login cannot remove both.
  • Deleting a backup needs different permissions from deleting the database.
  • Someone restores a backup into a clean environment on a schedule and checks the data. AWS's guidance is to recover data periodically to verify backup integrity, and to confirm that the restore fits the RTO and the data loss fits the RPO.
  • The restore time you measured goes into the runbook as the expected time.

If the product is moving to a new cloud, agree the targets before cutover. Our cloud migration checklist shows where recovery targets fit in that sequence.

Disaster recovery test plan: from tabletop to failover

A disaster recovery test plan lists which tests you run, how often, and what counts as a pass. NIST describes two kinds of exercise. A tabletop exercise is discussion only: a facilitator presents a scenario and the team talks through roles and decisions, without deploying any equipment. A functional exercise has people perform their duties in a simulated environment.

A small team can cover both with three tests:

  1. Restore test. Restore the newest backup of one data store into a clean environment. Record the time taken and check the data.
  2. Tabletop. Walk through one scenario with the plan open. Note every step where someone says "I think" or "probably".
  3. Failover test. If you run a standby, switch to it. AWS's guidance is to test failover regularly, to verify that it works and that RTO and RPO are met.

How often is your decision. NIST leaves the frequency to the organization and scales the depth of the exercise to the impact level of the system. Whatever you choose, write each result in the test log, failures included. A failed test costs far less than a failed recovery.

A disaster recovery plan checklist before you sign off

Use this disaster recovery plan checklist to review a finished draft. Each line is a question with a yes or no answer.

  • Does every system in scope have an RTO and an RPO, set by the person who owns the business impact?
  • Has a test met each target, with the date in the test log?
  • Are backups stored apart from the systems they protect, with separate delete permissions?
  • Does each system have a runbook that a second engineer has followed without help?
  • Is it clear who declares a disaster, and who steps in if that person is away?
  • Can the team reach the plan and each other if chat, email or single sign-on is down?
  • Do vendor contacts and support plans cover the hours you would need them?
  • Is there a next review date, with an owner?

Each "no" is an item for the work list. Start with the ones about backups, because every other line depends on them.

How we build disaster recovery with your team

We work with founders and teams who have an enterprise security questionnaire to answer, an investor review coming, or an outage they do not want to repeat. We start with the worksheet: your systems, the targets the business needs, and what your current backups can deliver. Then we close the gap, which may mean a change to backups, a deployment rebuilt in code or a standby, as part of cloud infrastructure consulting.

Pipelines, monitoring and on-call sit with our DevOps consulting services. If you are in an outage now, software project rescue is the faster route. Every engagement is under an NDA, and the IP is always yours. Each project is different, so we quote it after a short call.

Frequently asked questions

What is a disaster recovery plan template?

A disaster recovery plan template is a pre-structured document with the sections a recovery plan needs: scope, roles, recovery targets, backups, activation rules, runbooks and a test log. You replace the placeholders with your own systems and numbers. It saves writing time, but the plan only becomes real once a test has met its targets.

What is the difference between RTO and RPO?

RTO, the recovery time objective, is the longest acceptable delay between an interruption and the service being restored. RPO, the recovery point objective, is the longest acceptable time since the last data recovery point, which sets how much recent data you can lose. RTO shapes your restore process, and RPO shapes how often you back up.

Can a small business use this disaster recovery plan template?

Yes. A disaster recovery plan template for small business use needs the same sections as a large one, with fewer rows. A team with one database and one application can fit the plan on two pages. The parts not to skip are the recovery targets, the backup details and one restore test with its time recorded.

How is a disaster recovery plan different from a business continuity plan?

A business continuity plan covers how the business keeps operating during a disruption: people, premises, suppliers and processes. A disaster recovery plan is narrower and restores IT systems and data. NIST describes it as a plan that may support a business continuity plan by recovering the systems those processes depend on.

How often should you test a disaster recovery plan?

There is no universal number. NIST leaves the test frequency to the organization and scales the exercise to the impact level of the system, and AWS advises testing failover regularly. A workable habit for a small team is a restore test on a set schedule and after any change to the database or backup setup, with each result logged.

Is a database replica a backup?

No. A replica copies every change from the primary, usually within seconds, including a mistaken delete or a faulty migration. It protects against hardware failure, not against bad changes. A backup is a copy from an earlier point in time, kept apart from the primary, that you can restore after the mistake has happened.

Sources and update history