Academy/VCP-VCF 9.0 Administrator (2V0-17.25)/Multi-Instance Fleet Operations with VCF Operations (VCF 9.0)
This lab targets VCF 9.0

Multi-Instance Fleet Operations with VCF Operations (VCF 9.0)

VCF 9.0Advancedvcp-foundationarchitectvcdx⏱ 180 min

Source-checked against the VCF 9.0.1 documentation: Constructs of a VCF Private Cloud (Table 1), Fleet Management chapter (identity and access, certificates, passwords, tags, configuration drifts, linking vCenter instances, add/remove/move VCF Instances across fleets), Lifecycle Management (component classification, release types), Licensing, NSX Federation, VMware Cloud Foundation Instance Recovery. A VCF fleet is one or more VCF Instances managed by one VCF Operations instance (with the fleet management appliance and VCF Automation); there is no separate 'Fleet Manager' product and SDDC Manager remains in every instance. Rewritten for v8.2.2 (1 Oct 2026) after the 360-degree audit.

Objectives

  • Describe the VCF 9.0 hierarchy (private cloud > fleet > instance > domains) and the split between fleet-level management components (VCF Operations, fleet management, VCF Automation, Identity Broker, Operations for logs and networks) and instance-level core components (SDDC Manager, vCenter, NSX Manager, ESX)
  • Add an existing VCF Instance (5.x or later) to a VCF fleet: VCF integration in VCF Operations with a system-managed service account, management vCenter as a fleet management deployment target, and a collector (cloud proxy) in the new instance
  • Operate fleet-wide services from VCF Operations: certificates (configure CA, auto-renewal, non-disruptive updates), passwords (update and remediate local accounts), tags pushed to VCF domain vCenter instances, and configuration drift templates
  • Explain what is and is not fleet-wide: identity (one broker per set of instances), vCenter linking groups instead of ELM, licensing per vCenter, lifecycle per component and release type, and NSX Federation (not VCF Operations) for consistent DFW across instances
  • Monitor the fleet from VCF Operations: inventory per VCF Instance, Fleet Management > Tasks, certificate and password alerts, configuration drifts and license usage
  • Analyse fleet failure modes: VCF Operations or fleet management unavailable, a VCF Instance unreachable, a management domain lost (VCF Instance Recovery), and removing or moving an instance between fleets
  • Articulate the operational differences between VCF 5.2 (one SDDC Manager console per instance) and VCF 9.0 (fleet-level VCF Operations console over per-instance SDDC Managers)

Prerequisites

VCF Instance A deployed by the VCF Installer with VCF Operations and the fleet management appliance in its management domain (the fleet-level components). A second VCF Instance B (5.x or later, for example a second Holodeck instance or a VCF 5.2 pod) with its own SDDC Manager, management vCenter and NSX Manager, reachable over the network from VCF Operations. Backups configured as in vcp-admin-09.

Prior labs: holodeck-02, vcp-admin-09

Required skills:

  • VCF 9.0 architecture and the fleet / instance / domain hierarchy
  • VCF Operations console navigation (Administration > Integrations, Fleet Management, Inventory)
  • Certificate authority concepts (Microsoft CA templates, OpenSSL)
  • vCenter permissions and NSX roles
  • NSX Federation concepts (Global Manager, Local Managers)

Lab Environment

One VCF fleet. Instance A: management domain with SDDC Manager, vCenter 9.0, NSX Manager cluster, plus the fleet-level components VCF Operations, fleet management appliance and (optional) VCF Automation and VCF Identity Broker. Instance B: management domain with its own SDDC Manager, vCenter and NSX Manager, to be added to the fleet. One VCF Operations collector (cloud proxy) deployed into Instance B.

Mermaid source
graph TB
  subgraph Fleet[VCF fleet]
    OPS["VCF Operations + fleet management appliance (in Instance A mgmt domain)"]
    AUTO[VCF Automation]
    subgraph A[VCF Instance A]
      SMA[SDDC Manager A] --- VCA[vCenter A] --- NSXA[NSX Manager A]
    end
    subgraph B[VCF Instance B]
      SMB[SDDC Manager B] --- VCB[vCenter B] --- NSXB[NSX Manager B]
      CPB[Collector / cloud proxy]
    end
    OPS -->|VCF integration, system-managed service account| SMA
    OPS -->|VCF integration via collector| CPB --> SMB
  end

IP Addressing

NetworkPurposeVLAN
10.2.0.0/20Instance A management networkVLAN 1645
10.4.0.0/20Instance B management networkVLAN 1665

Credentials

SystemUsernamePassword
VCF Operationsadmin (local)Set at VCF Installer deployment of Instance A
SDDC Manager A / Badmin@localSet at each instance's deployment; used to create the VCF integration
vCenter A / Badministrator@vsphere.localSet at each instance's deployment
NSX Manager A / BadminSet at each instance's deployment

Tasks

Task 1 Map the fleet hierarchy and the fleet-level versus instance-level components

Manageability

The VCF 9.0 vocabulary is precise and the exams test it: a VCF private cloud contains one or more fleets; a fleet is managed by one set of fleet-level components (VCF Operations and VCF Automation) and contains one or more VCF Instances and possibly standalone vCenter instances; an instance consists of SDDC Manager, NSX, vCenter and ESX organised in a management domain and optional workload domains. The management domain of the first instance hosts the fleet-level components. Nothing in 9.0 removes SDDC Manager.

Step 1

In VCF Operations, open Inventory and expand the VCF Instance entries. For Instance A record: management domain name, SDDC Manager FQDN, management vCenter FQDN, NSX Manager cluster FQDN, workload domains. Then open Fleet Management > Lifecycle > VCF Management > Components and list the fleet-level components with their versions: VCF Operations, fleet management appliance, and any of VCF Automation, VCF Identity Broker, VCF Operations for logs, for networks, orchestrator.

Two lists: instance-level core components of Instance A, and the fleet-level management components, each with FQDN and version.
Step 2

Reproduce Table 1 'Constructs of a VCF Private Cloud' from memory and check it against the documentation: VCF Instance = SDDC Manager, NSX, vCenter, ESX, organised as a management domain plus optional workload domains; VCF fleet = one or more VCF Instances, possibly standalone vCenter instances, managed by VCF Operations and VCF Automation, with the first instance's management domain hosting those fleet components; VCF private cloud = one or more fleets. Then Table 2: the management domain is created by the VCF Installer and contains SDDC Manager and the fleet management appliance; workload domains are created in VCF Operations, each with its own vCenter deployed into the management domain and an NSX Manager cluster that is new or shared with another workload domain, never the management domain's.

The hierarchy written in your notebook with the component lists per construct.
Step 3

Log in to SDDC Manager A directly (https://<sddc_manager_a_fqdn>, admin@local). Observe the notice that the SDDC Manager UI is being deprecated and that lifecycle activities are performed from VCF Operations after upgrade to 9.0. Then open Inventory > Workload Domains and confirm the same domains VCF Operations shows. Record which operations the SDDC Manager UI still exposes and which are now done from VCF Operations (workload domain creation under Inventory > VCF Instance > Add Workload Domains; backup settings under Administration > SDDC Manager; passwords and certificates under Fleet Management).

A short table: operation, 5.2 console, 9.0 console.
Step 4

Verify instance autonomy: log in to vCenter A and NSX Manager A directly and confirm you can operate them without VCF Operations. Note the documented rule that the SDDC Manager UI accepts only local accounts once VCF SSO is configured, while the SDDC Manager API accepts SSO users.

Direct component access confirmed; the VCF Operations console is the management layer, not a dependency for the data plane or for the component UIs.
Step 5

Draw the fleet topology: fleet-level components in Instance A's management domain; Instance A and Instance B as peers, each with its own SDDC Manager, vCenter and NSX Manager; the planned collector in Instance B; the corporate identity provider feeding the identity broker; the SFTP backup target from vcp-admin-09.

Topology diagram saved for the design defence.

Validation Gate

Check: Instance-level and fleet-level component lists, the constructs tables, the 5.2-vs-9.0 console table and the topology diagram are complete.

Expected: You can name every component of a fleet and say at which level it operates.

Common Errors

A 'Fleet Manager' appliance is drawn as a separate orchestrator that replaced SDDC Manager
Cause: Confusion between the VCF Operations fleet management appliance (which adds workflows to VCF Operations) and SDDC Manager (still present per instance).
Fix: Draw VCF Operations plus the fleet management appliance at fleet level and one SDDC Manager inside each instance's management domain.
The management domain's NSX Manager is shown as shared with workload domains
Cause: Misreading of the NSX sharing rule.
Fix: Workload domains may share an NSX Manager cluster with another workload domain, never with the management domain (Table 2).
VCF Operations is treated as the thing that must be up for workloads to run
Cause: Management plane confused with data plane.
Fix: Workloads, vCenter, NSX forwarding and vSAN continue when VCF Operations is down; fleet-level operations (lifecycle, monitoring, automation intake) stop.

Task 2 Add VCF Instance B to the fleet

Manageability

A fleet grows by integrating an existing VCF Instance with the fleet's VCF Operations: a VCF integration (account) for the new instance using a system-managed service account, the instance's management vCenter registered as a fleet management deployment target, and a VCF Operations collector deployed into the new instance so data collection stays local to it. This is the documented procedure in 'Add, Remove, or Move VCF Instances across VCF Fleets'.

Step 1

Confirm the prerequisites: you have administrative access to the VCF Operations instance that manages the fleet, and Instance B runs VCF 5.x or later. Confirm network reachability from the VCF Operations network to SDDC Manager B, vCenter B and NSX Manager B (DNS forward and reverse, HTTPS).

Prerequisites confirmed; DNS and HTTPS reachability tested.
Step 2

In VCF Operations (fleet), navigate to Administration > Integrations and click Add. Under Account Types select VMware Cloud Foundation. Enter Name (alias of SDDC Manager B), Description, optional Physical Datacenter; under VCF Credentials enter SDDC Manager B's FQDN and a credential (create one if not present), choose the Default Collector Group for now, click Validate Connection, then Add. On the Integrations page open the VMware Cloud Foundation account, find the new endpoint, click its ellipsis and select Enable Collection. Then edit the integration to use a system-managed service account as 'Configuring a Standalone VCF Account in VCF Operations' describes.

Instance B appears as a VCF endpoint in VCF Operations with collection enabled and a system-managed service account.
Step 3

Register Instance B's management vCenter as a deployment target in fleet management: Fleet Management > Lifecycle > Settings > Deployment Targets > Add, select vCenter B and complete the wizard ('Add a Deployment Target and Import VCF Operations to VCF Operations fleet management').

vCenter B listed under Deployment Targets.
Step 4

Deploy a VCF Operations collector (cloud proxy) into Instance B's management domain from VCF Operations ('Deploy the VCF Operations collector'), then edit the Instance B VCF integration and move it from the Default Collector Group to the new collector or collector group.

Collector online in Instance B; the Instance B integration collects through it.
Step 5

Validate: in Inventory, both VCF Instances are listed with their domains, vCenters and NSX Managers; Fleet Management > Certificates > VCF Instances and Fleet Management > Passwords > VCF Instances show Instance B; Fleet Management > Tasks shows the integration tasks completed.

Instance B is a full member of the fleet across inventory, certificates and passwords.

Validation Gate

Check: VCF integration for Instance B created with collection enabled and a system-managed service account; vCenter B added as a deployment target; collector deployed and bound to the integration; Instance B visible in Inventory, Certificates and Passwords.

Expected: Two-instance fleet managed from one VCF Operations console.

Common Errors

Validate Connection fails when adding the VCF integration
Cause: DNS or certificate trust between VCF Operations and SDDC Manager B, or wrong credentials.
Fix: Fix forward and reverse DNS, accept the SDDC Manager certificate, and use an SDDC Manager admin credential; retry Validate Connection.
Collection works but metrics for Instance B are slow or incomplete
Cause: Instance B collected through the Default Collector Group across sites.
Fix: Deploy a collector in Instance B and bind the integration to it.
After removing an instance from a fleet, a new instance cannot be added
Cause: Stale record of the deleted instance in the fleet management appliance database.
Fix: Follow KB 424495 to remove the deleted VCF Instance from the fleet management appliance database before adding another instance.

Task 3 Operate fleet-wide services: certificates, passwords, tags and configuration drifts

Manageability

Fleet Management in VCF Operations centralises licences, identity and access, certificates, passwords, configurations and tags across the instances in a fleet. Each of these has documented scope rules (what is fleet-wide, what is per instance) and you must know them to design operations correctly.

Step 1

Certificates: in VCF Operations, Fleet Management > Certificates. Click VCF Management to view certificates of the fleet-level components (fleet management, VCF Automation, VCF Identity Broker, VCF Operations, for logs, for networks); click VCF Instances and each instance name to view ESX, NSX Manager, SDDC Manager and vCenter certificates (toggle Show ESX Hosts); click Certificate Alerts. Record expiry dates and alerts for both instances.

Certificate inventory for VCF Management and for Instance A and B with expiry and alert status.
Step 2

Configure a Certificate Authority: Fleet Management > Certificates > VCF Management > Configure CA (Microsoft CA only for the VCF management components: CA Server URL ending in certsrv, least-privileged service account, template name), then VCF Instances > Instance A > Configure CA (Microsoft CA or OpenSSL on the SDDC Manager appliance), and again for Instance B: the CA is configured separately for VCF Management and for each VCF Instance. Verify time synchronisation between the Microsoft CA and the SDDC Manager appliances.

A CA configured for VCF Management and for each instance; the fleet-level CA is Microsoft CA.
Step 3

Enable auto-renewal for the non-disruptive certificates listed in the documentation (ESX SSL, vCenter machine SSL, NSX LM and VIP, SDDC Manager SSL, VCF Identity Broker, VCF Operations, VCF Automation, for logs, for networks, fleet management) and generate a CSR and replace the certificate of one component in Instance B to exercise the workflow.

Auto-renewal enabled for the listed certificates; one certificate replaced from the console.
NSX supports non-disruptive rotation of the vCenter machine SSL certificate, but a Root CA rotation is disruptive: after it, edit and save the Compute Manager registration in NSX to import the new root.
Step 4

Passwords: Fleet Management > Passwords. Review the VCF Management accounts (VCF Operations root and admin; fleet management root and admin@local; for networks consoleuser, support, admin@local; for logs root and admin; VCF Automation admin; Identity Broker vmware-system-user) and the VCF Instance accounts (ESX root, vCenter root, SDDC Manager backup, NSX root/admin/audit). Check Fleet Management > Tasks for failed or running workflows, then Update the password of one account in Instance B (one at a time), following the recommended policy (15-20 characters, mixed classes, no dictionary word, palindrome, monotonic sequences or triple repeats). Then change a password directly in a component and use Remediate to resynchronise VCF.

One password updated and one remediated from VCF Operations; alerts page understood. Note: Rotate (auto-generated) remains an SDDC Manager action; expiry information in VCF Operations refreshes once a day (the SDDC Manager API is real time).
Step 5

Tags: Fleet Management > Tags. Create a category (for example Tenant, single tag per object, VM object type) with tags team-a and team-b, then Push Tag Categories to both VCF domain vCenter instances (version 9.0 required). Verify the category exists in vCenter A and vCenter B.

Category and tags pushed to both vCenter instances from one place; conflicts (same category name with different properties) reported if present.
Step 6

Configuration drifts: Fleet Management > Configuration Drifts. Create a configuration template from a vCenter 9.0 or vSphere cluster in Instance A (desired state), assign it to a policy covering the clusters of both instances, and run a drift check. Review the drifts reported for Instance B.

A template, a policy and a drift report across instances (applies only to vCenter instances and clusters deployed in VCF 9.0 or later).
Step 7

Licensing: open the licensing area of VCF Operations (Licensing Overview; connected or disconnected mode via the VCF Business Services console) and record how licences are added to VCF Operations and assigned per vCenter instance (primary and add-on licences). Note that licences are assigned to vCenter instances, not pushed as a fleet policy.

Licence inventory and the assignment model recorded.

Validation Gate

Check: Certificates viewed and CA configured at fleet and instance level with auto-renewal; one password updated and one remediated; tags pushed to both vCenters; configuration template and drift report across instances; licence model recorded.

Expected: Fleet-wide services exercised from VCF Operations with their documented scopes.

Common Errors

Configure CA offers no OpenSSL option for VCF Management
Cause: VCF management components support Microsoft CA only; OpenSSL is available for VCF Instance components.
Fix: Use a Microsoft CA for the fleet-level components; configure the instance CAs separately.
Password update fails or is blocked
Cause: A failed or running workflow exists, or two updates attempted at once.
Fix: Clear failed workflows under Fleet Management > Tasks, wait for running ones, and update one password at a time.
Tag push fails with conflicts
Cause: The target vCenter has a category with the same name but different properties.
Fix: Resolve the properties or select the overwrite option when pushing; remember categories set to Multiple cannot be changed to Single.

Task 4 Understand what is fleet-wide and what is not: identity, vCenter linking, lifecycle, licensing and NSX Federation

Manageability

Several things a designer might expect VCF Operations to push across instances are either scoped differently or belong to another product. Identity is per identity broker (one broker can serve up to five instances); the vSphere Client single pane of glass is vCenter linking in VCF Operations, not ELM; lifecycle is per component and release type; licences are per vCenter; consistent DFW across locations is NSX Federation, not a VCF Operations policy.

Step 1

Identity: in Fleet Management > Identity & Access > SSO Overview, view the identity broker and the instances connected to it. Connect Instance B to the existing SSO configuration (Enable Single Sign-On > Connect to an existing SSO configuration, select the identity broker, configure Instance B's vCenter and NSX as clients, acknowledge the role-assignment requirement), then assign roles in vCenter B and NSX B. Record the design limits: one identity provider per broker; the embedded broker recommended for one instance; up to five instances per appliance broker; VCF Operations and VCF Automation connect to one broker.

Instance B's vCenter and NSX connected to the fleet's identity broker with roles assigned; SSO scope and limits recorded.
Step 2

vCenter linking: Infrastructure Operations > Configurations > Logical Groupings > vCenter Linking > Create Group; add vCenter A and vCenter B (both 9.0, forward and reverse DNS, common identity provider users, not in ELM; 2 to 15 vCenters per group). Log in to vCenter A and confirm the vSphere Client shows the inventory of vCenter B as well. Note that from 9.0 vCenter linking replaces ELM for the single pane of glass, that IDP user permissions are per vCenter and must be set on each, and that a vCenter in ELM must have ELM deactivated first.

A vCenter linking group with both management vCenters; cross-vCenter inventory visible from one vSphere Client.
Step 3

Lifecycle: Fleet Management > Lifecycle. Open Binary Management and note how install, upgrade and patch binaries are downloaded (online depot or the VCF Download Tool for air-gapped environments); open the component list for VCF Management and the lifecycle view of each VCF Instance. Record the release types and consumption rules: major (9.0.0.0) and minor (9.1.0.0) releases are fully synchronised full-stack upgrades; maintenance releases (9.0.1.0) are synchronised with a new BOM but components can be applied selectively; patches (9.0.1.0100) are not synchronised and are applied per component. Then write how the core upgrade order (SDDC Manager, NSX, vCenter, ESX per domain) is driven from VCF Operations per instance.

Release-type table and a per-instance lifecycle plan; binaries managed once at fleet level.
Step 4

Licensing: confirm that licences are added to VCF Operations and assigned to each vCenter instance (primary licence, add-on licences) and that licence usage analytics are fleet-wide; there is no 'fleet licence policy' pushed to instances.

Licence assignment per vCenter recorded for both instances.
Step 5

Consistent firewall policy across instances: VCF Operations does not push DFW rules to several NSX Managers. Read the NSX Federation section (Global Manager, Local Managers, stretched segments, 'configure and enforce firewall rules consistently across locations', maximum 150 ms latency between locations). Document the two options for a fleet: NSX Federation with a Global Manager where the design needs one policy across locations, or per-instance NSX policy created in each Local Manager (manually or by automation) with the consistency risk recorded. Create the same DFW rule in NSX A and NSX B by hand as the non-Federation path: a rule in the Infrastructure category that allows the management subnet to the DNS servers group over DNS (UDP/TCP 53) and logs; verify in each NSX Manager under Security > Distributed Firewall.

Design note with both options and their constraints; an identical rule present in both NSX Managers.
Never write a DFW example that drops all traffic to DNS servers; it blocks name resolution for every workload. Scope deny rules to the specific untrusted sources and keep the default rule change (Allow to Drop) as the last step after the allow rules are verified.

Validation Gate

Check: Instance B connected to the fleet's SSO; vCenter linking group created; release types and lifecycle flow recorded; licensing model recorded; DFW consistency options documented with the hand-created rule in both NSX Managers.

Expected: You can state precisely which services VCF Operations provides fleet-wide and which remain per instance or belong to NSX Federation.

Common Errors

A design claims VCF Operations pushes DFW rules to all instances
Cause: Fleet management confused with NSX Federation.
Fix: Consistent firewall rules across locations are an NSX Federation capability; VCF Operations manages certificates, passwords, tags, configuration drifts, identity and lifecycle.
vCenter cannot be added to a linking group
Cause: It is below 9.0, in ELM, or already in another group.
Fix: Upgrade, deactivate ELM (cmsso-util break-elm or the SDDC Manager API), and check the compatible/incompatible lists in the wizard.
A patch is expected to upgrade the whole stack
Cause: Release-type confusion.
Fix: Patches and maintenance releases are selective per component; major and minor releases are full-stack, synchronised upgrades.

Task 5 Analyse fleet failure modes and instance recovery

Availability

The fleet introduces new dependencies: the fleet-level console, the collector per instance, the identity broker, and the fleet SFTP target. Each failure has a documented behaviour, and a lost management domain has a documented recovery (VMware Cloud Foundation Instance Recovery). This task makes you observe or document each, and practise removing an instance from the fleet.

Step 1

Build the failure-mode matrix before testing: VCF Operations and fleet management unavailable (what stops: fleet lifecycle, monitoring, identity management console, fleet SFTP backups of Automation and Identity Broker; what continues: workloads, vCenter, NSX forwarding, vSAN, SDDC Manager per instance, local logins; enterprise logins continue only if the identity broker is not the embedded broker of a vCenter that is also down); a VCF Instance unreachable from VCF Operations (collection stops for that instance; the instance operates); a collector lost (collection for its instance stops; redeploy the collector); a management domain lost (VCF Instance Recovery).

Failure-mode matrix with impact and continuity for each case.
Step 2

Test: power off the VCF Operations appliance(s) and the fleet management appliance in the lab. Verify from Instance B: vCenter B and NSX B UIs work with local accounts; workloads keep running; SDDC Manager B works; an enterprise login works or fails depending on where the identity broker runs. Power the appliances back on and confirm Fleet Management > Tasks and Inventory recover.

Observed: instance components and workloads unaffected; fleet-level functions paused and resumed.
Step 3

Test: block network traffic between VCF Operations and Instance B (lab firewall rule). Verify Instance B shows collection errors in Administration > Integrations while Instance B itself keeps operating. Remove the block and confirm collection resumes.

Collection outage isolated to Instance B; instance autonomy confirmed.
Step 4

Document the loss of a management domain using 'VMware Cloud Foundation Instance Recovery': failure scenarios (complete instance failure, ransomware, datastore corruption); the as-built documentation to keep (topology diagrams, vmnic to switch-port mappings, VMkernel IPs, DNS/NTP/AD); the recovery timeline for the management domain (deploy ESX hosts, deploy the VCF Installer, partial VCF deployment, deploy vCenter, first NSX Manager node and SDDC Manager from OVA, restore vCenter, NSX Manager and SDDC Manager, recover clusters, remediate NSX, recover NSX Edge nodes, post-recovery tasks, then VCF Operations and fleet management from image-based backups, Identity Broker and VCF Automation redeployed and restored, then for logs and for networks); and the VMware.CloudFoundation.InstanceRecovery PowerShell module prerequisites.

Instance recovery runbook outline for the management domain of a fleet-hosting instance.
Step 5

Remove Instance B from the fleet as documented (and, optionally, re-add it): in VCF Operations, Administration > Integrations > VMware Cloud Foundation > ellipsis on Instance B > Delete; Administration > Cloud Proxies > delete Instance B's collector; Fleet Management > Lifecycle > Settings > Deployment Targets > delete vCenter B. If VCF Automation consumed resources from Instance B, delete those objects in VCF Automation first. Before adding another instance to this fleet, follow KB 424495 to clean the deleted instance from the fleet management appliance database.

Instance B removed from the fleet; Inventory, Certificates and Passwords no longer list it; KB 424495 step recorded.
Step 6

Re-add Instance B (Task 2) if you want the fleet back, then write the design decision for your own environment: one fleet-wide identity broker versus cross-instance brokers; collector per instance; backup targets per instance and for the fleet; and the blast radius of the fleet-hosting instance's management domain.

Fleet design decisions with justification and implications in RCAR form.

Validation Gate

Check: Failure-mode matrix written and two failures observed in the lab; instance recovery runbook outline written; Instance B removed from the fleet by the documented steps.

Expected: Fleet dependencies and their failure behaviour are understood and recorded.

Common Errors

VCF Operations outage is reported as a workload outage
Cause: Management plane and data plane confused.
Fix: Workloads, vCenter, NSX forwarding and vSAN continue; only fleet-level operations stop.
Enterprise logins fail everywhere during a management vCenter maintenance
Cause: Embedded identity broker in that vCenter.
Fix: Expected for embedded mode; design the appliance broker for critical uptime, and keep local accounts available.
Cannot add a new instance after removing one
Cause: Stale entry in the fleet management appliance database.
Fix: Apply KB 424495 before adding the next instance.
Instance recovery attempted without as-built documentation
Cause: Topology and host network mappings not kept outside the platform.
Fix: Keep the documentation set the Instance Recovery guidance lists on secure secondary storage.

Final Validation

A two-instance VCF fleet is operated from one VCF Operations console: Instance B integrated with a collector and deployment target; certificates, passwords, tags and configuration drifts managed fleet-wide with their documented scopes; identity, vCenter linking, lifecycle, licensing and NSX Federation placed correctly; failure modes observed and the instance recovery runbook outlined.

✓ Hierarchy and component lists → Private cloud > fleet > instance > domains; fleet-level vs instance-level components correct; SDDC Manager per instance

✓ Instance B added → VCF integration with system-managed service account, deployment target, collector; visible in Inventory, Certificates, Passwords

✓ Certificates and passwords → CA per VCF Management and per instance; auto-renewal on; one password updated, one remediated

✓ Tags and configuration drifts → Category pushed to both vCenters; template, policy and drift report

✓ Scope rules → SSO connected for Instance B; vCenter linking group; release types; licences per vCenter; DFW consistency via NSX Federation or per-instance rules

✓ Failure modes and recovery → Matrix, two lab observations, instance recovery outline, instance removal by the documented steps

✓ 5.2 vs 9.0 articulation → Console table: per-instance SDDC Manager UI in 5.2; fleet-level VCF Operations over per-instance SDDC Managers in 9.0

Cleanup / Restore

Snapshot: vcp-admin-10-complete

• Re-add Instance B to the fleet if it was removed and you want to keep the two-instance lab

• Remove the lab firewall rule used for the isolation test

• Confirm VCF Operations, fleet management and the collector are powered on and collecting

• Keep the failure-mode matrix, the topology diagram and the recovery outline for the design defence

Design Reflection (VCDX)

A panel will ask you to place every management function at the right level. Fleet level (VCF Operations with the fleet management appliance, VCF Automation, Identity Broker, Operations for logs and networks): licences, identity and access, certificates, passwords, tags, configuration drifts, lifecycle binaries and orchestration, monitoring. Instance level: SDDC Manager, vCenter, NSX Manager and ESX, each still backed up and operated per instance.

Then the scope rules: one identity provider per identity broker and up to five instances per appliance broker; vCenter linking (2-15 vCenters, no ELM) instead of ELM; major and minor releases as synchronised full-stack upgrades, maintenance and patch releases selective; licences per vCenter; NSX Federation for consistent DFW across locations with the 150 ms latency limit.

Finally the dependency analysis: what stops and what continues when VCF Operations, a collector, the identity broker or a management domain is lost, and the Instance Recovery timeline for the management domain that hosts the fleet components. Saying that SDDC Manager was removed, or that a fleet console pushes firewall policy, would fail the defence.

Requirements

  • All VCF Instances of the fleet are integrated with the fleet's VCF Operations with a collector per instance
  • Certificates for VCF Management and for each instance are issued by a configured CA with auto-renewal for the non-disruptive set
  • Local account passwords are managed and remediated from VCF Operations
  • Tags and configuration templates are defined once and pushed or applied across instances
  • Identity is centralised with an identity broker whose scope matches the fleet design (embedded per instance or appliance for up to five instances)
  • Consistent network security across locations is delivered by NSX Federation where required
  • A documented instance recovery plan exists for the management domain hosting the fleet components

Constraints

  • VCF management components support Microsoft CA only; instance components support Microsoft CA or OpenSSL
  • Only one password update at a time; expiry data in VCF Operations refreshes daily
  • vCenter linking needs vCenter 9.0, no ELM, common IDP users, 2-15 members
  • One identity provider per identity broker; appliance broker recommended for up to five instances
  • NSX Federation requires at most 150 ms latency between locations
  • KB 424495 clean-up is required between removing and adding instances in a fleet

Assumptions

  • Instance A hosts the fleet-level components and is healthy
  • Instance B is VCF 5.x or later and reachable from VCF Operations
  • A Microsoft CA with a VMware template is available for the fleet-level certificates
  • Backups from vcp-admin-09 are in place before the failure tests
  • The lab firewall allows the isolation test without affecting other labs
  • Instance B's capacity can host a collector appliance

Risks

  • Loss of Instance A's management domain removes the fleet console — Impact: fleet operations stop until Instance Recovery completes — Mitigation: image-based backups of VCF Operations and fleet management, documented recovery timeline, as-built documentation off-platform
  • Embedded identity broker in a single vCenter — Impact: enterprise login outage during vCenter maintenance — Mitigation: appliance broker for critical uptime; local accounts under password management
  • Collector failure — Impact: monitoring blind spot for one instance — Mitigation: collector groups, alerting on collection status
  • Certificate expiry across many components — Impact: component outage — Mitigation: auto-renewal and certificate alerts in Fleet Management
  • Inconsistent DFW across instances without Federation — Impact: security drift — Mitigation: NSX Federation, or automation with periodic comparison
  • Stale fleet database after instance removal — Impact: cannot add instances — Mitigation: KB 424495 in the runbook

Self-Assessment Discussion Prompts

  1. Which functions move to fleet level in VCF 9.0 and which stay in each instance's SDDC Manager, vCenter and NSX Manager?
  2. When would you run one identity broker for the whole fleet versus one per set of instances, and what does each choice do to the blast radius?
  3. Why does VCF 9.0 recommend vCenter linking over Enhanced Linked Mode, and what has to happen to an upgraded 5.x instance before either SSO or linking works?
  4. How do major, minor, maintenance and patch releases differ in consumption, and how does that shape the maintenance calendar for a fleet?
  5. A customer wants one firewall policy across two VCF Instances in different cities; what do you propose and what are the constraints?
  6. What exactly stops when VCF Operations is down, and how do you prove to a panel that workloads are unaffected?
  7. How does the recovery of the fleet-hosting management domain differ from the recovery of any other instance's management domain?

Extensions

Cross-instance identity broker design

Deploy an appliance identity broker, connect Instances A and B to it, and document the fleet-wide versus cross-instance SSO models with their attributes (scale, resources, operational overhead, availability, blast radius, SSO scope) from the VCF Single Sign-On Detailed Design.

harder

NSX Federation in the lab

Deploy an NSX Global Manager, onboard NSX A and NSX B as Local Managers, stretch a segment and create one DFW policy spanning both locations; measure latency against the 150 ms limit.

harder

Air-gapped lifecycle

Use the VCF Download Tool to populate Binary Management offline and apply a maintenance release to one component in Instance B only.

same

Configuration templates under source control

Manage the configuration drift templates with a supported source control solution and run drift checks from a committed template.

same

⚠ Known Pitfalls

Inventing a 'Fleet Manager' product with policy engines COMMON
Problem: Designs describe a separate Fleet Manager with certificate, password, patching and DFW policies pushed to instances.
Resolution: The fleet-level console is VCF Operations with the fleet management appliance. It manages certificates, passwords, tags, configuration drifts, identity and lifecycle with documented scopes; DFW consistency is NSX Federation.
Declaring SDDC Manager removed in 9.0 COMMON
Problem: Backup, upgrade and recovery plans omit SDDC Manager.
Resolution: Every VCF Instance still has SDDC Manager in its management domain; its UI is deprecated in favour of VCF Operations, the appliance is not removed.
Reusing ELM for the single pane of glass COMMON
Problem: Upgraded instances keep ELM and cannot join VCF SSO or vCenter linking.
Resolution: Deactivate ELM (cmsso-util break-elm or SDDC Manager API, resolve drift) and use vCenter linking groups in VCF Operations.
Forgetting the fleet database clean-up after removing an instance COMMON
Problem: Adding a new instance fails after one was removed.
Resolution: Follow KB 424495 to remove the deleted instance from the fleet management appliance database.
Collecting every instance through the default collector group COMMON
Problem: Cross-site collection is slow and fragile.
Resolution: Deploy a VCF Operations collector in each instance and bind the integration to it.

References

Was this page useful?
Type to search. ↑ ↓ to move, Enter to open, Esc to close.