VMware ESXi Host Operations

Testbed Requirements

ESXi hosts must be defined in the pyATS testbed with os: linux (ESXi shell is Linux-based):

devices:
  esxi-host-01:
    os: linux
    type: esxi
    connections:
      cli:
        protocol: ssh
        ip: 10.0.0.100
        port: 22
    credentials:
      default:
        username: "%ENV{NETCLAW_USERNAME}"
        password: "%ENV{NETCLAW_PASSWORD}"

Note: SSH must be enabled on the ESXi host (disabled by default). Enable via vSphere Client → Host → Configure → System → Services → SSH → Start.

How to Call

All commands use pyats_run_linux_command:

PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-01","command":"<command>"}'

Commands

VM Inventory

List All Virtual Machines

PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-01","command":"vim-cmd vmsvc/getallvms"}'

Returns a table of all VMs on the ESXi host:

Column	Description
Vmid	Unique VM identifier (used as parameter for other vim-cmd operations)
Name	VM display name
File	Path to the VMX configuration file on the datastore
Guest OS	Configured guest OS type (e.g., ubuntu64Guest, windows9Server64Guest)
Version	Virtual hardware version (e.g., vmx-19, vmx-20)
Annotation	VM description/notes

What to check:

• All expected VMs are present and registered
• VMX file paths reference expected datastores
• Virtual hardware versions are current (old versions may lack features)
• Guest OS type matches actual installed OS
• No orphaned or unexpected VMs

Snapshot Management

Get VM Snapshots

PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-01","command":"vim-cmd vmsvc/snapshot.get 1"}'

The parameter is the Vmid from vim-cmd vmsvc/getallvms. Returns the snapshot tree for that VM:

• Snapshot name and description
• Snapshot ID
• Creation timestamp
• Whether it's the current snapshot (the "you are here" marker)
• Parent/child snapshot hierarchy

What to check:

• Stale snapshots — snapshots older than 72 hours degrade performance (growing delta disks)
• Snapshot chains — deep chains (>3 levels) cause I/O performance degradation
• Pre-change snapshots — verify a snapshot was taken before maintenance windows
• Disk space — snapshot delta files consume datastore space proportional to guest write activity

---

Workflows

1. VMware Inventory Audit

pyats_list_devices → identify ESXi hosts in testbed
→ vim-cmd vmsvc/getallvms per host → collect all VMs
→ Cross-reference with NetBox/Nautobot → flag undocumented VMs
→ Check virtual hardware versions → flag outdated versions
→ GAIT

2. Snapshot Health Check

pyats_list_devices → identify ESXi hosts
→ vim-cmd vmsvc/getallvms per host → get VM IDs
→ vim-cmd vmsvc/snapshot.get per VM → collect snapshot trees
→ Flag: snapshots > 72 hours old, chains > 3 deep, orphaned snapshots
→ Severity-sort (stale snapshots = WARNING, deep chains = CRITICAL)
→ GAIT

3. Pre-Change VM Verification

ServiceNow CR must be in Implement state
→ vim-cmd vmsvc/getallvms → identify target VM by name → get Vmid
→ vim-cmd vmsvc/snapshot.get {vmid} → verify pre-change snapshot exists
→ If no snapshot: STOP — snapshot must be taken before changes
→ Record baseline state in GAIT

4. Multi-Host VM Inventory

pyats_list_devices → identify all ESXi hosts
→ vim-cmd vmsvc/getallvms per host (parallel) → collect fleet-wide VM inventory
→ Aggregate: total VMs, VMs per host, hardware version distribution
→ Cross-reference with vCenter inventory (if available)
→ GAIT

---

Parallel Operations

Run VM inventory across multiple ESXi hosts concurrently:

# ESXi Host 1
PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-01","command":"vim-cmd vmsvc/getallvms"}'

# ESXi Host 2
PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-02","command":"vim-cmd vmsvc/getallvms"}'

# ESXi Host 3
PYATS_TESTBED_PATH=$PYATS_TESTBED_PATH python3 $MCP_CALL "python3 -u $PYATS_MCP_SCRIPT" pyats_run_linux_command '{"device_name":"esxi-host-03","command":"vim-cmd vmsvc/getallvms"}'

For snapshot checks, first collect VM IDs per host, then run snapshot queries in parallel across all VMs.

---

vim-cmd Quick Reference

Command	Description
vim-cmd vmsvc/getallvms	List all registered VMs (ID, name, file, OS, version)
vim-cmd vmsvc/snapshot.get {vmid}	Get snapshot tree for a specific VM
vim-cmd vmsvc/power.getstate {vmid}	Get VM power state (on/off/suspended)
vim-cmd vmsvc/get.summary {vmid}	Get VM summary (CPU, memory, tools status)

Note: Only getallvms and snapshot.get are covered by this skill. The reference above is for context — expand to other vim-cmd operations as needed.

---

Integration with Other Skills

Skill	Integration
pyats-linux-system	System-level commands on the ESXi host (ps, ls, docker stats)
pyats-linux-network	ESXi host networking (interfaces, routes) — useful for vmkernel adapter inspection
pyats-network	Network device show commands complement ESXi host-level views
pyats-parallel-ops	pCall pattern for fleet-wide ESXi operations
netbox-reconcile	Cross-reference VM inventory with NetBox virtualization records
nautobot-sot	Validate ESXi host and VM data against Nautobot
servicenow-change-workflow	Gate snapshot operations behind ServiceNow Change Requests
gait-session-tracking	Every vim-cmd execution logged in GAIT
nvd-cve	Check ESXi versions against NVD vulnerability database
cml-lab-lifecycle	CML labs may run on ESXi — correlate VM inventory with CML topology

---

Guardrails

• Always call pyats_list_devices first — verify ESXi hosts exist in the testbed
• Read-only operations only — getallvms and snapshot.get are read-only
• Never create/delete snapshots via this skill — snapshot creation/deletion impacts VM performance and requires a ServiceNow CR
• Never power-cycle VMs — vim-cmd vmsvc/power.* operations are destructive and require explicit authorization
• Validate Vmid before use — always get the Vmid from getallvms output; never guess or hardcode
• Flag stale snapshots — any snapshot > 72 hours should be flagged as WARNING; > 7 days as CRITICAL
• SSH access concerns — ESXi SSH is often disabled in production; document that it must be enabled for pyATS access
• Record in GAIT — every vim-cmd execution must be logged