Skip to content

Availability

Availability

The Container Service offers the same service level as the rest of the Cloud Service, check it here.

See below for additional details per orchestrator.

Kubernetes

Clusters are currently available both single master and multi-master.

Masters are at the moment spread across multiple nodes but a single AZ.

Node downtime should have the following impact on the cluster:

  • Master: On multi master cluster no impact unless a full AZ is unavailable. The latter case is similar to single master clusters, where the control plane becomes unavailable, no impact on running application pods
  • Nodes: Degraded capacity, no downtime for Services/Pods with appropriate number of replicas

Node loss should have the following impact on the cluster:

  • Master: Single master clusters need to be recreated, unless etcd_volume_size was set - cluster metadata backed by CEPH, check the Service Level Description of Ceph as this will impact the master. No impact for multi-master clusters
  • To create a single node cluster using etcd_volume_size, set the size in gigabytes: openstack coe cluster create --merge-labels --labels etcd_volume_size=5
  • For multi-master clusters see this section
  • Nodes: Lost cluster capacity (temporary), no downtime for Services/Pods with appropriate number of replicas

If possible, we recommend splitting production deployments into two or more separate clusters to ease upgrades and minimize the impact of cluster failures.


Last update: November 13, 2020