TechOps Workbench
ALWAYS ON · WINDOWS SERVER FAILOVER CLUSTERING

Count the votes.
Understand the failure boundary.

Explore a single failure event against an explicit pre-event vote snapshot.

Cluster vote snapshot

01 / INPUTS

“Vote” means the node’s effective vote immediately before the event. Use DynamicWeight from your cluster, not just configured NodeWeight. Availability describes the selected surviving partition after the event.

Witness availability means this partition can claim its vote, not merely reach the endpoint. A witness vote cannot be counted by both sides.

Site buttons change node availability only. Set witness availability independently to match its location and arbitration outcome.

◇ Scenarios stay in your browser.

Static majority result

02 / RESULTS
FROZEN VOTE SNAPSHOT · NOT A CLUSTER SIMULATOR

Votes available to the selected partition

—

Scenario summary

Read-only cluster inspection

Import-Module FailoverClusters
Get-ClusterQuorum
Get-ClusterNode |
  Select-Object Name, State, NodeWeight, DynamicWeight
Get-Cluster |
  Select-Object Name, DynamicQuorum, WitnessDynamicWeight

Run with cluster read access from a host with the FailoverClusters module. Commands inspect the cluster and do not change votes or force quorum.

UNDERSTAND THE BOUNDARY

Quorum is necessary.
It isn’t failover readiness.

WSFC votes belong to cluster nodes, not availability replicas or databases. Include every relevant cluster node, even nodes outside the selected availability group.

The static calculation

Total pre-event votes = entered effective node votes + entered effective witness vote. Required majority = floor(total ÷ 2) + 1. The selected partition has an arithmetic majority when its available votes reach that threshold and at least one node is available.

Unavailable votes remain in the denominator for this frozen snapshot. Do not remove them after a failure merely to make the calculation pass. All selected available nodes are assumed able to communicate in a single partition.

Dynamic quorum changes the picture

Windows can change effective node and witness votes as membership changes. Sequential failures can therefore produce a different result from a simultaneous loss. This tool does not simulate those transitions, tie-breaking, startup arbitration or witness locking. It cannot predict whether a live cluster will stay online.

Inspect DynamicWeight and WitnessDynamicWeight immediately before the event being modeled. A configured witness may currently have zero votes. Do not assume every NodeWeight = 1 node currently has an effective vote.

Always On has additional requirements

A quorum majority does not guarantee automatic AG failover. Replica health, synchronization, commit mode, failover mode, endpoint connectivity and failover policy still matter. Losing a site can affect a witness and network paths as well as nodes; model those separately.

Scope: Windows WSFC vote-snapshot planning. Not Linux/Pacemaker, distributed AG arbitration or a recovery procedure. Guidance checked October 5, 2026.