Installation Manual - 69 Old Vault VM power-off

Governed power-off of the old lost-custody Vault VMs after replacement Vault R1 validation.

This chapter records the governed power-off of the old lost-custody v7 Vault VMs after replacement Vault R1 had absorbed OpenShift consumers and post-stage 1 scheduled backups passed.

The old VM definitions and disk images were preserved. No disk image was deleted.

Governance

Preflight

The gate started only after the following conditions were true:

• stable vault.v7.comptech-lab.com resolved to replacement Vault R1;
• old main Vault DNS records were already absent;
• old seed DNS was preserved;
• hub/spoke External Secrets egress policies allowed only R1 Vault CIDRs;
• ExternalSecrets were 6/6 Ready on both clusters;
• OADP DPAs were Reconciled and BSLs were Available;
• post-stage-1 scheduled backups had passed;
• R1 Vault health returned HTTP 200 on all three R1 main nodes.

The normal scheduled backup windows also ran after the earlier accelerated validation:

One hub-side OCM wrapper Application for spoke reported Synced/Progressing, but the MulticlusterApplicationSetReport showed the managed spoke app as healthy:

clusters=1 synced=1 healthy=1 inProgress=0 notHealthy=0 notSynced=0

The spoke-local Argo application was Synced/Healthy.

Power-Off Action

Old VMs targeted:

• gf-ocp-vault-01
• gf-ocp-vault-02
• gf-ocp-vault-03
• gf-ocp-vault-seed-01

Actions:

1. disabled libvirt autostart for the four old Vault VMs;
2. requested graceful ACPI shutdown for the three old main VMs and the old seed VM;
3. waited until all four old VMs reported shut off.

No forced virsh destroy was needed.

VM State After Power-Off

Old disk images were retained:

DNS And Vault Health

DNS after power-off:

PowerDNS zone serial remained:

44

Old Vault direct health after power-off:

R1 Vault health after power-off:

OpenShift Validation

Argo CD:

OADP:

External Secrets:

Vault egress policies still allowed only R1 Vault CIDRs:

30.30.200.35/32,30.30.200.36/32,30.30.200.37/32

Cluster state:

• both clusters remained OpenShift 4.20.18;
• all nodes were Ready;
• no cluster-operator exceptions were reported.

RHACS:

• hub Central remained Available;
• no non-running StackRox pods were reported on hub or spoke.

Result

The old lost-custody Vault VMs were powered off cleanly and will not autostart on the next hypervisor boot.

Replacement Vault R1 remained healthy, OpenShift consumers remained healthy, OADP backups remained valid, and the old disk images were retained.

The next gate should be a cold-retention soak. Disk deletion remains a separate final retention decision.