[cinder] Pass terminationGracePeriodSeconds to CinderNetappConfig CR for KVM pods

[hemna]

The kvm-cinder-netapp-operator CRD already supports the terminationgraceperiodseconds field (verified live in qa-de-1), but the Helm chart was not populating it. KVM cinder-volume pods were getting the Kubernetes default of 30s, while VMware volume and backup pods get 900s.

Problem

Measured in qa-de-1:

Pod type	terminationGracePeriodSeconds
VMware volume (cinder-volume-vmware-vc-*)	900
VMware backup (cinder-volume-backup-vmware-vc-*)	900
KVM volume (cinder-volume-kvm-*)	30

With 30s, any in-flight KVM volume operation taking longer than 30 seconds (e.g., volume create from image, migration) will be SIGKILLed by Kubernetes during pod shutdown.

Fix

One line added to cindernetappconfig.yaml to pass pod.terminationGracePeriodSeconds.volume into the CinderNetappConfig CR:

terminationgraceperiodseconds: {{ .Values.pod.terminationGracePeriodSeconds.volume }}

The value defaults to 30 in values.yaml and is overridden to 900 per region in secrets (e.g., qa-de-1/values/cinder.yaml already sets pod.terminationGracePeriodSeconds.volume: 900).

No secrets change needed

The secrets already set volume: 900 — the value just wasn't being passed to the operator CR.

Required for

Graceful shutdown feature: sapcc/cinder#358

Related

• [cinder] add --single-child to dumb-init for backup pods #12685 — adds --single-child to dumb-init for backup pods (same graceful shutdown prerequisite)
• cinder: use dumb-init --single-child for graceful shutdown #11680 — added dumb-init --single-child for VMware volume pods