Skip to content

Build1 publisher2 min readPublished

Every cross-account role in an AWS landing zone stops at the aws-eusc partition line

AWS's landing zone guidance for the European Sovereign Cloud starts at the partition boundary. Credentials in aws-eusc cannot reach resources in aws, so the centralisation an operator already runs gets built a second time inside the new partition.

The Engineer · Build desk

Illustration accompanying Every cross-account role in an AWS landing zone stops at the aws-eusc partition line

What happened

  • AWS published landing zone guidance for the European Sovereign Cloud, which it describes as a new independent cloud for Europe, physically and logically separate from existing AWS Regions and operated inside the EU.
  • The EUSC runs as its own partition, aws-eusc, with a separate control plane, IAM, billing, console and service endpoints, while offering the same services, features and APIs as commercial Regions.
  • Its first Region is eusc-de-east-1 in Brandenburg, Germany, and aws-eusc joins aws, aws-cn and aws-us-gov on AWS's list of partitions.
  • Cross-Region features including Amazon S3 Cross-Region Replication and Transit Gateway inter-Region peering work only between Regions that sit in the same partition.
  • The guidance ships Terraform and CloudFormation snippets that resolve the partition at deploy time so that ARNs are never hardcoded with the arn:aws prefix.

Compiled by The EngineerSomething wrong?How this is made

Why it matters

  • cost Every centralised component an operator built once gets funded and staffed twice, and the second log pipeline into the SIEM is a running cost.
  • constraint Disaster recovery out of Brandenburg into a commercial Region cannot use the platform replication features, so it becomes an application-level copy job driven with two credential sets.
  • decision Any module that spells out arn:aws has to move to a partition-derived ARN before it can be reused, and teams already deploying into aws-us-gov have written that code once already.
  • capability Workloads that must run on EU-operated infrastructure can target the same service APIs, so application code ports and the rebuild is confined to the operating plane.

Every AWS Region belongs to exactly one partition, each partition runs its own instance of IAM, and AWS describes the boundary between partitions as hard [3]. A landing zone is mostly trust relationships. They all stop at that line. Credentials issued in one partition cannot be used against resources in another [4]. The role an identity account assumes in a workload account has no cross-partition equivalent, because caller and callee are registered in different IAM instances.

For the case AWS spells out, a CI/CD system in a commercial Region deploying into EUSC, the guidance is to integrate at the network or API layer with separate credentials for each partition [8]. Cross-partition trust is not on offer [8]. The pipeline ends up holding two credential sets and calling eusc-de-east-1 endpoints as an external client.

The post says one organization, one logging account, one identity source and one billing roll-up are each achievable within a partition, and that EUSC gets an independent landing zone mirroring the commercial operating model [7]. Four centralised components in each of two partitions is eight [14].

The code change is small. It lands in every module that ever wrote an ARN out by hand. AWS's Terraform and CloudFormation examples build the ARN from a partition value resolved at deploy time, and the guidance is not to hardcode arn:aws [9]. The same pattern is presented as working unchanged across aws, aws-us-gov and aws-eusc [9].

The most useful thing in the post is its own labelling. It marks which behaviors are platform boundaries of a sovereign partition and which are configuration choices you can adapt, and maps the design to the AWS Security Reference Architecture and the Well-Architected Framework [11]. Billing roll-ups, SSO, cross-account roles, Direct Connect and image distribution are the questions AWS says the boundary settles, with the practical differences set out in a table [10]. A companion post covers alignment with C5:2020 criteria and an independent assessment report on AWS Artifact [12].

How much of that mirror is worth building depends on how much runs in aws-eusc. The model assumes an operator wants the commercial operating model reproduced inside aws-eusc [7]. For a two-account EU estate, most of that structure is overhead carried so the audit story reads the same on both sides. For an estate that has to stay inside the EU, the duplication is the cost of the boundary AWS drew, and AWS describes that boundary as hard [3].

What to watch

  • A second EUSC Region would put S3 Cross-Region Replication and Transit Gateway inter-Region peering back in play for in-partition DR.
  • The rows of AWS's practical-differences table for Direct Connect and image distribution decide how much network work the bridge actually takes.
  • Whether the C5:2020 assessment report and compliance workbook on AWS Artifact are accepted as covering aws-eusc deployments.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories