# IaaS FAQ

Frequently asked questions about Sunrise Cloud Services, compute, memory, storage, GPUs, Capacity Pools, Namespaces, and Hosted Clusters.

For product documentation, see [Sunrise Cloud Services](../../iaas/index.md).

---

## Questions

==- Where is my data physically located?
Your data resides entirely in Switzerland. Our data centre is located in Basel and is fully managed by Sunrise, with no third-party operator handling your infrastructure. Backups are held separately at Mount10, a high-security facility in the Swiss Alps. This keeps both your live workloads and your backup copies under Swiss jurisdiction at all times, which is a key requirement for organisations bound by data-residency and sovereignty mandates.

==- Are GiB the same as GB?
The Sunrise Portal displays and bills memory and storage in GB. The capacity you actually receive is the same number in GiB. If you purchase 10 GB, you are allocated 10 GiB. Because 1 GiB is larger than 1 GB, the delivered capacity is slightly greater than the GB figure suggests, never less.

==- How do I keep different teams or environments isolated from each other?
Isolation happens at the Namespace level. A Namespace is a governed slice of a Capacity Pool with its own resource quotas and access controls, so multiple teams or environments can share the same pool while staying separated through ResourceQuota limits and RBAC policies. A common pattern is one Capacity Pool with separate dev, staging, and production Namespaces. If you need stronger separation than shared-cluster multi-tenancy provides, a dedicated Hosted Cluster is the alternative.

==- What GPUs do you offer and how are they allocated?
The current offerings are NVIDIA H100 (80GB) and H200 (141GB), purchased as subscriptions and assigned to a Capacity Pool like any other resource. GPUs are allocated as whole units and are not shareable across Namespaces: unlike CPU, a GPU assigned to one workload cannot be partially consumed by another. This matters for capacity planning, since you cannot oversubscribe a single GPU the way you might with compute. They are used for AI inference and other accelerated workloads.

==- What is the vCPU limit per VM?
The maximum CPU allocation for a single VM is **55 vCPUs**. Any VM you provision on the platform cannot be sized beyond 55 vCPUs.

==- When should I use a Namespace versus a Hosted Cluster?
Run in a Namespace when you want workloads on Sunrise' shared, managed clusters with quota and access controls handling separation. Choose a Hosted Cluster when you need a dedicated OpenShift cluster with full administrative access and your own control plane, while Sunrise manages the underlying control-plane infrastructure. A Hosted Cluster gives you a dedicated etcd, API server, and scheduler, worker nodes sized to your needs, a stable cluster API hostname, and Ceph-backed storage.

==- My Hosted Cluster finished provisioning but doesn't work, what should I check?
Even when credentials and the **Launch OpenShift Console** link appear in the portal, the console and API may still need a short time to become fully available. Wait a minute or two and try again.

If the cluster still does not respond, confirm that you provisioned the minimum required storage for the Hosted Cluster. Worker root volumes (50 GB per worker minimum), etcd storage (10 GB per member minimum), and total capacity pool storage as described in the [Hosted Cluster sizing guide](../../iaas/platform-architecture/hosted-cluster.md#sizing-and-configuration). Insufficient storage is a common cause of clusters that provision but fail to run correctly.

==- How does failover and disaster recovery work?
In-region failover between availability zones is included at no extra cost. Multi-region disaster recovery (geographically distinct) is available for an additional cost.

High availability (HA) requires parallel, dedicated resources, billed at the same rate as primary workloads. Architecture and HA planning (for example, active/active) can be delivered as a consulting engagement at CHF 1,500/day, depending on complexity.

==- What SLA does Sunrise provide?
Sunrise commits to a minimum Uptime percentage of 99.99% for Sunrise Online Services, measured on a 24x7x365 basis, excluding scheduled maintenance agreed in advance ("Service Availability Rate").

==- Is the Ceph RGW S3 endpoint externally reachable (public Route / LoadBalancer)?
Yes. Sunrise exposes an external endpoint for Ceph RGW S3: [https://ocs-storagecluster-cephobjectstore-secure-openshift-storage.apps.basel.kvant.cloud/](https://ocs-storagecluster-cephobjectstore-secure-openshift-storage.apps.basel.kvant.cloud/). Use this URL when configuring S3 clients or applications that need to reach object storage from outside the cluster. Access still requires valid S3 credentials; the public endpoint does not grant access on its own.

==- Can backups be stored outside the primary region?
Yes. Backups can be archived to out-of-region object storage hosted by [Mount10](https://mount10.ch/en/about-us/swiss-fort-knox/) to support long-term retention and geographic isolation for disaster recovery.

==- Does Sunrise help design backup strategies?
Yes. Sunrise can assist with backup architecture and provide the required infrastructure. More complex scenarios are supported via consulting at CHF 1,500/day.

==- What is Flux Status and what does it show me?
Flux Status is a web UI for the GitOps engine that runs your workloads. Sunrise clusters reconcile your deployments from a Git source, and Flux Status shows you what is currently running, what is failing, and the error behind each failure, in real time. See [Monitor GitOps State](../../iaas/monitor-gitops-state.md) for full documentation.

==- How do I get access to Flux Status?
Each cluster has its own URL: [flux.basel.phoeniqs.com](https://flux.basel.phoeniqs.com) for Basel, and [flux.ai-3.phoeniqs.com](https://flux.ai-3.phoeniqs.com) for AI-3. Log in with your Sunrise SSO credentials. You will only see the namespaces your account has been granted access to.
===