B.1 PodDisruptionBudget Kubernetes Resource
Background
Pods are the basic units of Kubernetes and are used to deploy Cloud Native Core (CNC) Network Functions (NFs).
Pod disruption occurs due to voluntary and involuntary disruptions.
Voluntary Disruptions
Voluntary pod disruption occurs when:
- pods are removed from a cluster.
- SEPP worker nodes are drained at the time of cluster upgrade and running pods are evicted from these nodes.
- load balancing of worker nodes is performed by evenly distributing pods among worker nodes.
Involuntary Disruptions
Involuntary pod disruption occurs when:
- the system hardware or software fails.
- SEPP worker nodes are removed accidentally.
- pods are evicted due to insufficient resources on SEPP worker nodes.
Kubernetes has introduced the PodDisruptionBudgets (PDB) concept to manage pod disruption outages in the clusters. PDB is applicable only for voluntary disruptions. It is used to avoid service impact due to voluntary disruption and ensures high availability of NF.
PodDisruptionBudget (PDB) is a Kubernetes resource that allows you to achieve high availability of scalable application services when the cluster administrators perform voluntary disruptions to manage the cluster nodes. PDB restricts the number of pods that are down simultaneously from voluntary disruptions. Defining PDB is helpful to keep the services running undisrupted when a pod is deleted accidentally or deliberately. PDB can be defined for highly available and scalable OC SEPP services such as cn32f-svc, pn32f-svc, cn32c-svc, pn32c-svc, n32-ingress-gateway, n32-egress-gateway, plmn-ingress-gateway, plmn-egress-gateway, nfmanagement-svc, nfdiscovery-svc and perf-info services.
It allows safe eviction of pods when a Kubernetes node is drained to perform
maintenance on the node. It uses the maxUnavailable parameter specified in the Helm chart to
determine the maximum number of pods that can be unavailable during a voluntary disruption.
For example, when a maxUnavailable is 25%, pod evictions are allowed as
long as no more than 25% of the desired pods are in unhealthy state.
For more information about this functionality, see https://kubernetes.io/docs/concepts/workloads/pods/disruptions/#pod-disruption-budgets.
Default PodDisruptionBudget for SEPP Deployment
Table -2 Default PodDisruptionBudget for SEPP Deployment
| Microservice | PDB Value Specified (maxUnavailable) |
|---|---|
| cn32c-svc | 25% |
| pn32c-svc | 25% |
| cn32f-svc | 25% |
| pn32f-svc | 25% |
| config-mgr-svc | Not Applicable |
| plmn-ingress-gateway | 25% |
| plmn-egress-gateway | 25% |
| n32-ingress-gateway | 25% |
| n32-egress-gateway | 25% |
| app-info | 25% |
| config-server | 50% |
| perf-info | 25% |
| nf-mediation | 25% |
| coherence-svc | 25% |