Taking Back the Keys: ERP Migration from Vendor Servers to Client-Owned Kubernetes
Hosting, and therefore the data, sat with the vendor, and leaving was hard by design. We rebuilt the environment as code in the client's repo, migrated in parallel to containers on their own infrastructure, and handed over runbooks.
- Service
- Operations & Support
- Industry
- Enterprise IT & Cloud
- Region
- Germany
- Engagement
- Lab-type ODC (dedicated team)
Outcomes
- Full infrastructure & data ownership
- Environment as code
- Zero-surprise cutover with rollback
Stack
Odoo · Docker · Kubernetes · Terraform · PostgreSQL · CI/CD
Written byOriental Nippon Team
A German enterprise ran its core operational systems, including Odoo, on servers owned and operated by its software vendor. It was convenient until it wasn't. The client could not decide where its data lived, when its systems were updated, or how they were configured.
Leaving was hard by design. There was no infrastructure definition the client owned, so its environment could not be reproduced anywhere else.
The problem
- Hosting and data with the vendor. The client did not control either.
- Releases on someone else's schedule. Timing and configuration were out of the client's hands.
- Lock-in. Moving away was hard by construction.
- Nothing reproducible. The environment existed only on running servers.
| Before | After | |
|---|---|---|
| Hosting | Vendor servers | Client on-premise and cloud |
| Environment definition | None | Terraform in the client's repository |
| Deployment | Vendor process | Client-owned CI/CD pipelines |
| Runtime | Virtual machines | Docker on Kubernetes |
What we built
Migrated production instances
Moved to the client's own on-premise and cloud environments, run in parallel until each was proven.
Infrastructure as code
The environment is a Terraform definition in the client's repository.
Automated deployment pipelines
Releases are the client's decision and a routine operation.
Containers
From virtual machines to orchestrated containers, which is what made the environments portable.
Each production instance went through the same cutover states:
Technology and the role it played
| Technology | Role in this project |
|---|---|
| Odoo | The operational system being moved |
| Docker | Packaging services as containers |
| Kubernetes | Running those containers on-premise and in cloud |
| Terraform | Defining every environment as code in the client's repository |
| PostgreSQL | Operational data, migrated and verified |
| CI/CD | Client-controlled build, test and deployment |
How we delivered it
Much of the existing environment had never been written down. The first job was simply finding out what was running and what depended on it.
- Inventory. What runs, how it is configured, and what depends on it.
- Build targets as code. On-premise and cloud, before moving anything.
- Migrate in parallel. The vendor-hosted system stayed live until the new one was proven.
- Cut over. With a tested path back.
- Hand over operations. Runbooks, monitoring, backup and recovery, all owned by the client.
01Inventory and assessment
2–3 wks
Current environment documented, risks graded
02Target environments as code
3–4 wks
On-prem and cloud, containers, orchestration
03Pipelines
2–3 wks
Automated build, test and deployment
04Parallel migration
3–5 wks
Data and services moved, verified against the source
05Cutover and hand-over
2 wks
Switch-over, rollback tested, runbooks and monitoring
Risks and how we handled them
| Risk | How we handled it |
|---|---|
| Undocumented dependencies | Full inventory of the existing environment first |
| Downtime during the move | Parallel migration, with the vendor system live until proven |
| A cutover with no way back | A tested rollback path before switch-over |
| Lock-in to a new party | Everything handed over: code, runbooks, monitoring, recovery |
Outcome
The client controls its infrastructure and its data, on environments defined in its own repository that can be rebuilt, moved or scaled without anyone's permission.
Reliability, maintainability and portability all improved because the environment became a definition instead of a machine. That is the real deliverable, more than the move itself.
What the client keeps
- Infrastructure definitions, in their own repository
- Deployment pipelines
- Runbooks for every recurring operation
- Monitoring and backup configuration
- A tested recovery procedure
Sources and further reading
- What is Terraform?, HashiCorp
- Kubernetes overview, Kubernetes
- Docker overview, Docker
- Blue-green deployment, Martin Fowler
Facing something similar? This is the kind of work our Operations & Support team delivers every day. Talk to us about your stack.
