Running on Scaleway?

Scaleway instance or Kapsule misbehaving? We ship it properly.

A Kapsule cluster upgrade got stuck half way through, an Instance in Paris-1 stopped responding after an overnight maintenance notice, and the Object Storage bucket policy that worked last quarter now returns AccessDenied on a service account that should still have write access. We walk the Scaleway API, read the resource events, and bring the estate back to a state you can describe without opening the console.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

scaleway-api.log

still broken

What it is

Scaleway at a glance

Scaleway is a French cloud provider owned by the Iliad group, best known for its bare-metal roots as Online.net and for the newer product line built around Instances, Managed Kubernetes (Kapsule), Object Storage, Serverless Containers and Functions, Managed Databases, and Apple Silicon machines for iOS build farms. It operates datacenters in Paris, Amsterdam and Warsaw, with a sales focus on European customers and an API-first product style aimed at developers.

Scaleway is usually the European cloud that developers picked without being told to, where OVHcloud is often the one procurement chose. The product surface is coherent, the Terraform provider is first-party, and the console is modern. The trouble in production is usually around quotas that start low on new projects, Kapsule upgrades that stall on specific node pool conditions, Object Storage regions that are not interchangeable, RPN private networking that was not attached to new resources, and bare-metal or dedicated GPU servers whose provisioning takes longer than the surrounding tooling expects.

Type
European cloud with VMs, Kubernetes, bare metal, Object Storage and Serverless
First released
1999 as Online.net, 2015 as Scaleway
License
Proprietary service, open-source API clients and Terraform provider
Maintained by
Scaleway SAS (Iliad group)
Typically used for
European cloud hosting, developer-first Kubernetes, bare-metal workloads, S3-compatible storage, Apple Silicon build runners

The honest version

Scaleway pros and cons

No technology is all upside. This is where Scaleway earns its place and where it tends to cause trouble.

Where Scaleway is strong

  • A developer-first product surfaceThe API, the CLI and the Terraform provider cover almost everything the console does, and they feel like they were written by people who have used a cloud from the inside.
  • European data residencyDatacenters in France, Netherlands and Poland, under French corporate ownership, which clears the usual GDPR and sovereignty questions without a long legal diligence round.
  • Apple Silicon machinesScaleway rents real Mac mini hardware by the hour, which gives iOS teams a build farm without buying racks of MacBook Pros or paying an MDM provider to manage them.
  • Honest Managed KubernetesKapsule is a managed Kubernetes product with a sensible default node pool model, VPC-level networking and a reasonable price per node, suitable for small and medium clusters without a hyperscaler contract.
  • S3-compatible Object Storage with cold tiersStandard, Glacier-equivalent and cold storage all speak the S3 API, which makes Scaleway a credible backup and archive target without rewriting clients.
  • Transparent pricingPer-hour billing with monthly caps, clear bandwidth rules and no surprise control-plane charges. The invoice reads cleanly next to AWS or GCP.

Where Scaleway bites

  • Regional footprint is smaller than the hyperscalersThree regions, with one or two zones each, is enough for European-centric workloads and not enough for a global multi-region strategy. Latency to North America and Asia still goes through the public internet.
  • Object Storage regions are not interchangeableA bucket in pl-waw-1 is not the same namespace as pl-waw-2, and migrations between regions are an export-and-import job rather than a replication toggle.
  • Quotas on new projects are tightFresh projects start with low compute and storage quotas, and raising them is a support interaction rather than a self-service click, which surprises teams during a scaling moment.
  • Kapsule upgrades occasionally stallNode pool upgrades sometimes stick with a subset of nodes on the target version, usually due to pod disruption budgets or local-storage pods that cannot be drained. The console does not always make the cause obvious.
  • Managed Databases are fine, not deepManaged PostgreSQL, MySQL and Redis exist and work well for straightforward workloads. For high-availability setups, complex replication topologies or large datasets, teams still end up self-managing on Instances or dedicated servers.
  • Support depth varies by tierThe free support tier is competent and ticket-based. Engineers accustomed to hyperscaler enterprise support should budget for a paid tier if the workload needs fast response at odd hours.

Sound familiar?

Common Scaleway problems we fix

The Scaleway problems that bring people to us, and what you have once each one is fixed.

quota exceeded: Compute Dev1-S

Instance quotas are blocking a scale event

A project whose compute quota was never raised beyond the default, hitting a ceiling during a deploy that should have expanded the fleet. Terraform fails mid-run and leaves partial state.

Quotas raised and committed to Terraform

Kapsule: node pool upgrade stuck

A Kapsule upgrade has stalled

Pods with PDBs that will not allow a drain, local storage on nodes that is not being migrated, or an autoscaler argument during the rolling update. Half the cluster is on the new version and half is not.

Drains and PDBs understood, upgrade completed

403 AccessDenied: s3:PutObject

Object Storage permissions drifted

A service account whose IAM policy no longer matches the bucket policy after a rotation, or a client still pointing at an old endpoint that silently returns AccessDenied for anything outside the correct region.

Policies and endpoints aligned with current access

RPN v2: new instance not attached

New instances are missing the private network

A private network that was set up for the original fleet and not attached to instances created afterward, so services communicate over public interfaces that then hit firewall rules.

Private network attachments enforced

outbound SMTP port 25 blocked

Transactional mail stops leaving the instance

A production deploy that worked in staging because staging used a relay. Scaleway blocks port 25 by default on new resources, and the message queue grows silently.

Mail routed through a relay

NoSuchBucket on cross-region access

Object Storage requests hit the wrong region

A backup job that moved regions in configuration but not in endpoint URL, so the SDK is calling pl-waw-2 against a bucket that still lives in pl-waw-1. The error message is honest and easy to misread.

Region-aware clients and bucket paths

Flexible IP: unassigned after replacement

A flexible IP got left behind

An instance that was replaced through Terraform or the console, with the flexible IP unbound and not reassigned. The bill keeps running and the DNS record points at nothing.

Flexible IPs reconciled against live instances

Managed Database: maintenance window 03:00 CET

A Managed Database restart fell in a bad window

A maintenance window announced in CET and accepted without anyone noticing it maps to US business hours. The restart lands during live traffic rather than in the quiet period the team assumed.

Windows routed to real quiet hours

What we do

Scaleway help, from rescue to day-to-day

Ring us for one failing instance, or let us look after the whole Scaleway estate.

Rescue
Instances that stopped responding, Kapsule upgrades that stalled, or Object Storage policies that suddenly refuse writes. We stabilize first and explain afterward.
Build-out
A new Scaleway estate in Terraform with RPN private networks, Object Storage, Kapsule and Managed Databases wired together, documented and reviewable.
Kapsule and container platforms
Managed Kubernetes clusters sized, upgraded, autoscaled and wired to Object Storage and the Container Registry with the add-ons you actually need.
Migrations
Moving workloads from AWS, GCP or Azure onto Scaleway for cost or sovereignty reasons, with staged database replication and DNS cutover windows planned.
Apple Silicon build farms
iOS CI on Scaleway's Mac mini fleet, provisioned and scaled through the API so you are not paying for an idle rack of laptops.
Managed Tech
Patching, backup verification, cost reviews and someone to page when a Paris-1 zone has an unexpectedly busy night.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Scaleway alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Scaleway questions

Something else? Call+1 737 727 7327

Is Scaleway reliable enough for production?

For most European-centric workloads, yes. Paris, Amsterdam and Warsaw have been running for years, the API is stable, and Kapsule clusters behave well once the node pool shape is reasonable. The honest limits are a smaller regional footprint than the hyperscalers and shallower managed services, which teams either compensate for with a second provider or build around on bare metal.

Why can we not scale our Scaleway project past a certain point?

Usually a quota. Compute instances, block storage, floating IPs and Kapsule clusters all have per-project quotas that begin low and rise on request. The console shows the limit, and a quota increase is a support interaction rather than a self-service toggle. Teams that run into this during a scaling event file the ticket early for next time, which stops a second surprise.

How do we handle a stuck Kapsule upgrade?

First we identify which pods are blocking the drain. Pod disruption budgets that allow zero unavailable, local storage that cannot be moved, or stateful workloads without a proper eviction path are the usual suspects. Once the blocker is addressed, the upgrade continues from where it stalled; sometimes we also tune the surge and unavailability settings on the node pool so the next upgrade does not repeat the pattern.

Can you take over a Scaleway account from a previous developer?

Yes. We walk the organization transfer, rotate every API key, move billing onto a company card, and inventory what is actually running across Instances, Kapsule, Object Storage and Managed Databases so nothing drops off silently during the handover.

Can you move us from Scaleway to AWS, or the other way?

Both directions are routine. From AWS to Scaleway we usually see cost or sovereignty drivers; from Scaleway to AWS we usually see a need for deeper managed services or a global footprint. Either way we build the destination in Terraform, replicate the database in the background, cut DNS over during a quiet period, and hold the source in standby mode for a week or two before decommissioning.

How much does Scaleway help cost?

We quote for free. Describe which projects and regions are involved and what is going wrong, and we come back with a fixed price before any resources are touched.

Free estimate

Scaleway giving you trouble?

Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.