Files
cds-ai/docs/cirrus-data-cloud/esxvm-guide.md

2.4 KiB

Cirrus Data Cloud ESX VM Guide

This file covers read-only review and approval rules for direct vCenter VM hardware changes.

Scope

  • Use this guide for direct vCenter VM hardware work such as:
  • creating a new VM disk
  • attaching existing datastore VMDKs
  • assigning PCI passthrough devices

vCenter Access

  • Always log into vCenter 192.168.0.201.
  • Source /home/aw/code/cds/.env.credentials.local and use VCENTER_USER plus VCENTER_PASSWORD.
  • Standard CLI path: /home/aw/.local/bin/govc
  • Default datacenter: CDSHQ-Eng

Review-First Rule

  • Treat direct VM hardware changes as a two-step workflow:
  • first gather and report findings
  • then wait for explicit operator approval before making any change
  • Never create, attach, remove, or reassign hardware until the operator explicitly approves after seeing the summary.

VM Identification Rule

  • When the requested VM name is not present exactly as given, search for the closest live inventory name and report the mismatch before planning any change.
  • Do not act on a guessed VM silently.

Host And Datastore Checks

  • Find the specified VM and verify the ESXi host it is currently running on.
  • Default expected ESXi host is 192.168.1.165, but always verify live placement before planning changes.
  • Always identify and report the datastore where the VM is stored before planning disk work.

New Disk Creation Rule

  • When asked to create a new disk for a VM:
  • inspect the VM's current controller layout first
  • prefer the existing SCSI controller already backing the VM disks unless the operator explicitly asks for a different controller
  • report the controller label, controller type, and the next free unit number before proposing the change
  • report the target datastore and requested size before proposing the change
  • if the VM has snapshots, include that fact in the review summary before the operator approves

Existing Disk And FC Assignment Rule

  • For existing-disk and FC-passthrough requests, use the workflow in vm-lookup-and-assignment.md.
  • If the specified devices or disk files cannot be found exactly as requested, do nothing.

Summary Requirement

  • Before any direct VM hardware change, always report:
  • VM name found
  • guest hostname when available
  • ESXi host
  • datastore
  • current controller layout relevant to the request
  • snapshot state when relevant
  • exactly what would be changed
  • any blockers that mean nothing will be changed