Back to blog
awseu-south-2disaster-recoverydata-residencycloud-architecture

AWS Spain region (eu-south-2) or a second region: how to decide without overbuilding

How to choose between eu-south-2, your current European region and a second recovery region based on data residency, available services and real RTO/RPO needs.

“We want the data in Spain.” “If the region goes down, we lose the service.” “It will be faster from Spain.” In conversations about AWS, these three statements often appear together, as though they were a single decision. They are not. The first is a question of contracts and regulation, the second of recovery objectives, and the third is answered by measuring, not by looking at a map.

Confusing them leads to two opposite mistakes: migrating to eu-south-2 in the belief that this alone ensures GDPR compliance or protects against a regional outage, or building an architecture active in two regions that a small team cannot operate or test, and which therefore provides less protection than it promises. This article separates the three questions and proposes a decision-making sequence you can defend to a client, an auditor or your own board.

Three motivations, three different answers

What you are asked for or worried aboutWhat actually addresses itWhat does not address it
“The data must be in Spain / in the EU”Choosing the region (or regions) where each service stores and processes data, and documenting this in the data processing agreement and record of processing activitiesDeploying in Spain without reviewing global services, subprocessors or specific transfers
“The service must not go down”Multi-AZ within one region as a baseline, plus a cross-region recovery strategy sized around RTO/RPO for each workloadBeing in Spain: a region is still a single regional failure domain
“It must be fast from Spain”Measuring actual latency from users and integrations, including the CDN, the user's network and global servicesAssuming an improvement from geographical proximity without measuring it

If you can answer only one question this month, answer the first in writing: it is the one that usually comes from a contract, tender specification or due diligence exercise, and it shapes the others.

What exactly is eu-south-2?

eu-south-2 is the AWS Europe (Spain) region. According to the official AWS regions documentation, it has three Availability Zones (AZs) and is an opt-in region: regions launched after March 20, 2019 must be enabled in each account before they can be used. Enabling it is a minor administrative step, but it needs to be included in the plan: it affects every account in the organisation and may interact with your service control policies (SCPs) and quotas.

Do not confuse an AZ with a region. An AZ is an isolated partition within a region, with independent infrastructure; AWS states that each region has at least three AZs, and in Spain they are eus2-az1, eus2-az2 and eus2-az3, as listed in the Availability Zones documentation. Deploying across several AZs in eu-south-2 protects against a data centre or AZ failure, but not against a problem affecting the entire region. That distinction underpins everything else: multi-region is not a substitute for Multi-AZ, but a further step worth paying for only if the business requires it.

As for latency: having users in Spain makes an improvement over Ireland or Frankfurt plausible, but distance is only one of several factors — the user's internet provider network, the use of CloudFront or another CDN, global services that do not “live” in your region, and the actual pattern of calls between components may dominate the outcome. AWS itself includes proximity among the region selection criteria in the documentation cited above, alongside available services and legal requirements. Treat it as a hypothesis to measure with your actual users and integrations before justifying a migration on performance grounds. You will not find promised millisecond figures in this article: without measurements on your workload, they would be invented numbers.

AWS's position, in its data privacy FAQ, is that the customer chooses the region or regions where their content is stored, and AWS will not move or replicate it outside those selected without their agreement, except where necessary to provide services the customer has initiated or to comply with the law or a binding governmental order. This is a reasonable basis for residency, but its boundaries need to be reviewed service by service.

The AWS EU data protection page qualifies this: content remains in the selected region, but a small number of services involve data transfers — for example, to provide the service itself or unless an option is disabled — so it recommends reviewing each service's privacy features and carrying out the relevant transfer assessment. There are also global services (IAM, Organizations, Route 53 and CloudFront, among others), which by design are not confined to one region.

The practical conclusion for a buyer in Spain:

  • Residency in eu-south-2 or another European region answers “where is the data?”, not “are we GDPR-compliant?”. You still need to consider the Article 28 data processing agreement, records of processing activities, security measures, the list of subprocessors and the transfer assessment for each service.
  • “Data in the EU” and “data in Spain” are different requirements. If a client or tender specification requires Spain, picking a region elsewhere in Europe is not enough. If it requires the EU, Spain is one of several options, and the decision can be based on services and operations.
  • The AEPD's guide for cloud service customers — marked as obsolete by the AEPD and useful only as a due diligence framework, not as a statement of current law — already called for analysing the location of processing and data, subcontracting and international transfers before signing a contract. The final legal decision belongs to your DPO or legal adviser, not to this article or your platform team.

So document exactly what the contract or applicable regulation requires — a country, the EU or just the EEA — review it service by service, and put the answer in writing. “We are in Spain” is not a compliance answer; at most, it is part of the evidence.

If this issue arises from an audit, it is worth reading it alongside the checklist for preparing for a cloud audit or due diligence exercise, which covers how to present evidence without improvising.

Before falling in love with a region: which services are available there?

eu-south-2 does not necessarily offer the same catalogue as eu-west-1, or the same features within a service that is available. AWS says this plainly: some services may not be available in every region, and it maintains the official list of services by region precisely so you can check, including feature notes within each service.

Before deciding to migrate, do this boring but inexpensive work:

  1. Inventory the services your workload uses today and those your roadmap needs in the near term.
  2. For each one, check the official list to confirm that it is available in eu-south-2, along with the specific feature you use, not just the service name.
  3. Identify which components are global and what data they handle outside the selected region.
  4. Check that the region has been opted into in all affected accounts, along with quotas and organisational policies.

The catalogue changes over time, so the valid statement is “verified against the official list on the decision date”, never a definitive list of services that are present or absent. If an essential service is not available in Spain, the real options are to stay in the current region, deploy that component elsewhere and accept the hybrid cross-region architecture — with its latency and failure surface — or wait. None is free; choose knowing which cost you are taking on.

Account separation, incidentally, is a different axis from region selection: you can have production and development in separate accounts within the same region. If you have not addressed that axis yet, we cover it in separating production and development in AWS with multiple accounts; do not mix it into this decision.

A second region is justified by RTO/RPO, not fear

Once residency and services are settled, the recovery question remains. Here, the decision-making tool is not fear of a headline about a regional outage, but two objectives that the business must sign off for each workload, as required by Well-Architected Framework practice REL13-BP01:

  • RTO: the maximum acceptable time between disruption and service restoration.
  • RPO: the maximum tolerable data loss, measured as time since the last recovery point.

The same practice warns against arbitrary objectives or ones stricter than the business needs: they make the architecture more expensive without adding value. There is no single “company RTO”; there is one for the store generating revenue, one for the internal back office and one for the monthly report, and they are usually very different.

With RTO and RPO defined per workload, choosing between strategies is almost mechanical. AWS orders four strategies (REL13-BP02) by increasing cost and complexity, and decreasing RTO/RPO. The ranges below are the official guidance from that practice, not an AWS SLA or a commitment about your system:

StrategyThe idea in one lineOfficial RPO/RTO guidance
Backup and restoreBackups replicated to the second region; infrastructure is redeployed during recoveryRPO in hours; RTO up to 24 hours
Pilot lightReplicated data and a minimal core ready; the rest is deployed during recoveryRPO in minutes; RTO in tens of minutes
Warm standbyA scaled-down but functional version always running in the second regionRPO in seconds; RTO in minutes
Multi-region active/activeBoth regions serve traffic; the business decides how to resolve write conflictsRPO near zero; RTO potentially zero

Three limits worth keeping clear:

  • RTO ends when the service is restored in the recovery region. Up to that point, detection, the decision — often involving someone on call — execution and validation with the business do count. Failback (returning to the primary region, once the event has subsided) comes afterwards and is not part of that RTO: it must be designed and tested with its own objectives and controls, because after failover the databases in the recovery region are live and up to date and the primary must be resynchronised without losing or duplicating data. That the return does not extend the incident's RTO does not mean it is optional; it means you must not mix two different clocks into the same objective.
  • Continuous replication does not protect against corruption or deletion, because it replicates the damage too. You need point-in-time backups and actual restoration tests; we develop that protocol in what to require from an AWS backup recovery test, so we will not repeat it here.
  • A second region that is “on” incurs continuous charges; backup and restore and pilot light reduce that ongoing cost in exchange for a higher RTO. Without a calculator using dated assumptions, any cost figure would be invented: compare cost structures, not promised euro amounts.

The AWS disaster recovery whitepaper adds the decisive operational detail for small teams: even backup and restore requires redeploying infrastructure, configuration and code in the target region — which requires reproducible infrastructure as code — pilot light requires deploying changes to both regions, and active/active is the most complex and expensive option. Complexity that is not tested regularly tends to fail on the day it is needed.

A five-question decision framework

With that in mind, the decision comes down to answering these questions in writing:

  1. What does the contract or regulation require in writing about data location: Spain, the EU or nothing specific?
  2. Where are the users and integrations, and what latency has been measured — not assumed — from them?
  3. Which essential services and features are missing from eu-south-2 today, according to the official list consulted on this date?
  4. What RTO and RPO has the business signed off for each critical workload, and which strategy on the ladder meets them?
  5. Who operates failover and failback, who tests them and how often?

The usual answers from a small team in Spain tend to fall into one of these three scenarios:

  • One well-built region, with a recovery destination that meets residency requirements. The two cases call for different answers. If the written requirement is the EU (not a specific country), the service catalogue fits and the agreed RTO is measured in hours, run production across several AZs in one region — often eu-south-2 if residency and services align — replicate point-in-time backups to another eligible region in the EU, and test restores. If the requirement is strictly Spain-only, that second European region does not satisfy it: according to the official AWS table, eu-south-2 is the only AWS region located in Spain. Before promising that architecture, document exactly which data and backups the obligation covers, which services and backups fall outside its scope or are covered by an express contractual or other lawful exception, and where recovery will take place. If no valid exception covers backups outside Spain, choosing “another European region” does not resolve the conflict: keep the backups subject to the obligation within Spain too (which makes them less useful if the region itself fails), renegotiate the scope, or document that you cannot meet both strict Spain-only residency and recovery in another AWS region. This is the default scenario only if the recovery destination is permitted; it requires less ongoing work than keeping two regions active.
  • The current primary plus a recovery region. If you already operate in eu-west-1 and it works, or if Spain lacks a necessary service, you can keep your current primary and use another region for backups or pilot light. If residency requires it, use Spain as the primary instead. What matters is not which region sounds better, but whether the location of production, backups and restoration meets the applicable obligation in each case. Under a strict Spain-only requirement, using Spain only as the backup destination does not make production hosted abroad compliant, and using another European region as the recovery destination does not by itself meet the national requirement. Document production, backups and restoration separately, including any exceptions. Change the recovery design and contractual documentation, not necessarily the entire platform.
  • Warm standby or active/active, depending on the workload's objectives. These are different tiers for different targets. Choose warm standby when the business has signed off on RPO in seconds and RTO in minutes for that workload, you have the capacity to operate and scale it, and your tests support the choice in line with REL13-BP02. A scaled-down but functional version runs continuously in the second region. Reserve multi-region active/active for workloads where the agreed objective is RPO near zero and RTO potentially zero, traffic must be served from multiple regions, and the team can resolve write conflicts and test both regions continuously. Business approval and paying for the capacity are necessary, but do not replace that distinction: an RTO in minutes can justify warm standby without targeting near-zero recovery time, and warm standby must not be sold as a guarantee of near-zero RTO.

No scenario is “the right one” in the abstract. The right one answers the five questions with evidence, not geographical preferences.

What to validate before migrating or signing

Before committing to a date with a client or starting the migration:

  • Enable eu-south-2 in the affected accounts and check quotas and policies.
  • Finalise the service/feature inventory against the official list, recording the consultation date.
  • Review cross-region dependencies: a component that depends on another region's control or data plane may weaken precisely the recovery posture you are seeking; the workload must be able to operate in a single region.
  • Make sure infrastructure as code deploys to the target region without undocumented manual steps.
  • Test restoration or failover with stopping criteria and a return plan defined before starting, not during the test.
  • Put in writing which part of the answer is residency, which is recovery and which is measured performance. That page is worth more in due diligence than any untested multi-region diagram.

If this exercise reveals that you do not know where your platform stands, the preliminary step is not to migrate: it is to review. In Cloud Reliability, we work specifically on the reliability and operations of platforms already in production, including backups and recovery.

In closing: answer the question you have, not the one that sounds better

eu-south-2 is a good answer when residency in Spain is required in writing and the service catalogue fits. A second region is worth buying when agreed RTO/RPO requires it and there is a team to operate and test it. Latency is a good reason when it has been measured. Everything else is paying for complexity in exchange for imagined peace of mind — or rejecting a suitable region in favour of a multi-region setup that will never be operated properly.

If you need to decide on region, residency or recovery level before migrating or signing with a client, we address this through AWS cloud design and architecture: alternatives compared with their assumptions, structural costs and operational burden, and a phased plan your team can actually maintain. No promise of outcomes outside our control: the final decision, with its limits documented, will be yours and will be defensible.

Sources

AR

Alejandro Rodríguez

Freelance Cloud & DevOps Consultant, Madrid