Upcloud’s managed Kubernetes offers solid performance once past initial networking hurdles, but expect hands-on troubleshooting and limited upgrade flexibility.

What is Upcloud Managed Kubernetes?

Upcloud, a subsidiary of Updog, provides a managed Kubernetes service focused on European data centers. It abstracts control plane management but requires hands-on node and network configuration.

Operational Reality: Setup Pain, Post-Resolution Stability

Initial Setup Challenges:

• Internal networking misconfigurations (e.g., CNI plugin conflicts, subnet overlaps) caused multi-day outages during cluster creation.
• Documentation gaps required trial-and-error for VPC peering and DNS integration.
• Support response times varied: 24–48 hours for critical issues.

Post-Resolution Stability:

• Once networking stabilized, clusters ran with <1% downtime over 12 months.
• Resource pricing competitive with OVH/Exoscale, but no auto-scaling for system components.

Actionable Workflow: Setup and Troubleshooting Steps

1. Pre-Setup:

   • Define subnets and VPC peering before cluster creation.
   • Validate DNS delegation (e.g., coredns configmaps) pre-deployment.

2. Cluster Creation:

   • Use CLI for reproducibility:

     upcloud cli: cluster create --name prod-eu --region fi-abc --node-count 3  
     

   • Monitor node readiness:

     kubectl get nodes --watch  
     

3. Post-Creation Checks:

   • Verify CNI plugin status:

     kubectl -n kube-system get pods -l k8s-app=coredns  
     

   • Test external connectivity:

     curl -v http://<external-service-ip>  
     

4. Troubleshooting Networking:

   • Check for IP conflicts in Upcloud’s web UI under “Networks”.
   • Inspect kube-router logs for failed ARP entries.

Policy Example: Maintenance Window Enforcement

Policy:

• All control plane upgrades occur during EU business hours (09:00–17:00 CET).
• Node OS updates require 72-hour advance notice to Upcloud support.

Implementation:

• Use kubectl drain before maintenance:

  kubectl drain <node-name> --ignore-daemonsets --delete-emptydir-data  
  

Tooling and Integration

• CLI: Primary interface for cluster lifecycle management.
• Monitoring: Prometheus + Grafana with Upcloud’s built-in metrics endpoint.
• CI/CD: Helm charts for deployments; ArgoCD for sync.
• Limitation: No native integration for Velero backups—requires manual config.

Tradeoffs and Caveats

• Pros:
  • Strong data sovereignty compliance (EU-only data centers).
  • Predictable pricing with no hidden egress fees.
• Cons:
  • No automated node OS upgrades—manual intervention required.
  • Kube version upgrades lag upstream by ~3 months.

Troubleshooting Common Failures

1. Node Not Ready:

   • Check Upcloud’s hardware status in the console.
   • Reboot node via CLI:

     upcloud server reboot <server-id>  
     

2. DNS Resolution Failures:

   • Validate coredns deployment has correct forward zones.
   • Test with nslookup inside a debug pod:

     nslookup kube-dns.kube-system.svc.cluster.local  
     

3. Upgrade Failures:

   • Pre-upgrade: Drain nodes and backup etcd.
   • Post-upgrade: Monitor API server latency with Prometheus.

Final Note: Upcloud works for stable, low-touch workloads but demands operational discipline for setup and maintenance. For faster upgrades or auto-scaling, consider self-managed on their IaaS.

Source thread: Experience with Upcloud managed Kubernetes?