Skip to content

Windows-Cluster-Setup.md

Latest commit

 

History

History
239 lines (166 loc) · 10.6 KB

Windows-Cluster-Setup.md

File metadata and controls

239 lines (166 loc) · 10.6 KB
Raw

Windows Failover Cluster Setup

Environment context: These procedures were developed for building Windows Server Failover Clusters (WSFC) to support SQL Server Failover Cluster Instances (FCI) on Windows Server 2016+. The guidance focuses on the infrastructure preparation — disk layout, BIOS tuning, network configuration, and cluster creation — that precedes the SQL Server installation.

Table of Contents

Disk Recommendations

All disks should be initialized with GPT partition type to avoid the 2TB size limitation of MBR.

A mapped disk scheme is recommended for manageability — it uses a single drive letter mount point and is easier to maintain than assigning individual drive letters as disk count grows. The drive letter scheme below is a good starting point if mapped disks aren't feasible. The key is consistency across all servers and environments.

Purpose Drive Label Block Size Suggested Size Notes
Programs E:\ Programs 4KB (default) 450GB SSD SQL Server binaries and tools
Scripts/Files F:\ FILE01 4KB (default) 20GB Maintenance scripts, output files
Data H:\ DATA01 64KB 700GB SQL data files (.mdf, .ndf)
Logs I:\ LOG01 64KB 250GB SQL log files (.ldf)
TempDB T:\ TEMPDB01 64KB 250GB TempDB data files only; TempDB log goes on the Log disk
Backups X:\ BACKUP01 4KB (default) 250GB SQL backup files
Quorum Q:\ Quorum 4KB (default) 2GB Cluster quorum witness

The 64KB block size for Data, Log, and TempDB disks aligns with SQL Server's extent size (8 pages × 8KB = 64KB). See [[../Performance/Performance-Practices#disk-configurations|Performance Practices — Disk Configurations]] for a detailed explanation of why this matters.

Verify disk configuration after setup:

Test-DbaDiskAllocation -ComputerName ClusterNode01 | Select-Object * | Out-GridView

Server Hardware Preparation

These BIOS and hardware settings should be configured on all cluster nodes before building the cluster.

Disable Automatic System Recovery (ASR) in the RBSU/BIOS settings. ASR can clear diagnostic data needed to determine the root cause of a failure — you want the crash dump preserved for analysis, not auto-cleared on reboot.

Set CPU Power Management to Maximum Performance. Power-saving modes introduce latency variability that can affect SQL Server response times and cluster heartbeat reliability.

Set the BIOS clock source to local system time. The OS should manage time synchronization via NTP/Windows Time Service, not the BIOS. Mixed time sources can cause cluster validation warnings and log correlation headaches.

Active Directory Preparation

Pre-create the following objects in the target Active Directory OU before running the cluster creation wizard:

  • Cluster Name Object (CNO) — the computer account for the cluster itself.
  • Virtual Computer Objects (VCO) — one for each clustered role (e.g., the SQL Server FCI network name).

Set the CNO as the owner of the VCOs, then disable the VCOs. The cluster creation process will enable them when it takes ownership. Pre-creating and pre-permissioning these objects avoids the most common cluster creation failures related to AD permissions.

# Verify the CNO and VCO exist and are in the expected OU
Get-ADComputer -Identity 'ClusterName' -Properties DistinguishedName, Enabled
Get-ADComputer -Identity 'SqlFciName' -Properties DistinguishedName, Enabled

Enable the Failover Clustering Feature

The Failover Clustering feature must be installed on every node that will participate in the cluster.

# Install on a single node (2 params — inline)
Install-WindowsFeature -Name Failover-Clustering -IncludeManagementTools

# Install on multiple nodes remotely
$nodes = @('ClusterNode01', 'ClusterNode02')
foreach ($node in $nodes) {
    $splatFeature = @{
        Name                   = 'Failover-Clustering'
        IncludeManagementTools = $true
        ComputerName           = $node
    }
    Install-WindowsFeature @splatFeature
}

A reboot may be required after installation.

Create the Windows Failover Cluster

Validate the Configuration

Always run the cluster validation wizard before creating the cluster. This catches hardware, network, and storage issues that would cause problems later.

# Run all validation tests
Test-Cluster -Node ClusterNode01, ClusterNode02 -Verbose

# Review the validation report (HTML) — output path is displayed in the console

It's normal for the validation to return warnings, especially if shared storage is not yet presented or if you're using Storage Spaces Direct. Errors must be resolved before proceeding. Warnings should be reviewed and understood — the report will indicate if any are blocking.

Create the Cluster

$splatCluster = @{
    Name          = 'ClusterName'
    Node          = @('ClusterNode01', 'ClusterNode02')
    StaticAddress = '10.0.0.50'
    NoStorage     = $true
}
New-Cluster @splatCluster
  • StaticAddress — the IP address for the cluster management endpoint. Use an IP on the LAN/client network, not the heartbeat network.
  • NoStorage — skips automatic storage import. Add storage deliberately after creation to maintain control over disk assignments.

Ensure the DNS and Active Directory entries for the cluster name resolve correctly after creation.

Configure the Cluster Quorum

The quorum configuration determines how the cluster maintains consensus when nodes fail. The default is Node Majority, which works for two-node clusters with a witness.

For a two-node cluster, use a file share witness or cloud witness (Azure blob storage, available on Windows Server 2016+):

# File share witness (2 params — inline)
Set-ClusterQuorum -NodeAndFileShareMajority '\\FileServer01\ClusterWitness'

# Cloud witness (Azure)
$splatQuorum = @{
    CloudWitness = $true
    AccountName  = 'storageaccountname'
    AccessKey    = 'storageaccountkey'
    Endpoint     = 'core.windows.net'
}
Set-ClusterQuorum @splatQuorum

If using a disk witness instead (requires shared storage):

# The disk must already be added to the cluster (2 params — inline)
Set-ClusterQuorum -NodeAndDiskMajority 'Cluster Disk 1'

Rename Cluster Resources

Rename shared storage resources and network adapters before installing SQL Server. Default names like "Cluster Disk 1" and "Cluster Network 1" are meaningless during troubleshooting at 2 AM.

Shared Disks

Rename disks to match their purpose. The disk number in Failover Cluster Manager corresponds to the disk number in Disk Management — verify you're renaming the correct disk, especially when multiple disks share the same size.

# List current cluster disks
Get-ClusterResource | Where-Object { $_.ResourceType -eq 'Physical Disk' } |
    Select-Object Name, State, OwnerGroup

# Rename cluster disks
(Get-ClusterResource 'Cluster Disk 1').Name = 'DATA01'
(Get-ClusterResource 'Cluster Disk 2').Name = 'LOG01'
(Get-ClusterResource 'Cluster Disk 3').Name = 'TEMPDB01'
(Get-ClusterResource 'Cluster Disk 4').Name = 'BACKUP01'

Cluster Networks

Rename network adapters to identify their purpose. The LAN adapter (Cluster and Client) carries client traffic. The heartbeat adapter (Cluster Only) carries intra-cluster communication.

# List current cluster networks
Get-ClusterNetwork | Select-Object Name, Role, Address

# Rename
(Get-ClusterNetwork 'Cluster Network 1').Name = 'LAN'
(Get-ClusterNetwork 'Cluster Network 2').Name = 'Heartbeat'

Network roles:

  • Cluster and Client (Role = 3) — LAN adapter. Carries client connections and cluster traffic.
  • Cluster Only (Role = 1) — Heartbeat adapter. Carries only intra-cluster communication.
  • None (Role = 0) — Excluded from cluster use (e.g., backup network, iSCSI).

Post-Cluster Configuration

These settings should be applied on all cluster nodes after the cluster is built.

Disable Volume Shadow Copy (VSS) if not actively used. VSS snapshots on cluster shared disks can cause unexpected I/O latency and disk space consumption.

Set-Service -Name VSS -StartupType Disabled -ComputerName ClusterNode01
Set-Service -Name VSS -StartupType Disabled -ComputerName ClusterNode02

Configure the page file to a fixed size. Auto-managed page files can grow unexpectedly and consume disk space on the OS drive. A fixed size of 30GB is a common starting point for servers with 128GB+ RAM, but the right size depends on your memory dump configuration.

Set the OS to optimize for background services. Under Advanced System Settings → Performance → Advanced tab → Processor Scheduling, select "Background services."

Configure memory dump to Full Dump. Under Advanced System Settings → Startup and Recovery, set the dump type to Complete Memory Dump. Kernel dumps are smaller but may not capture enough state for cluster-related failures.

Verify firewall rules. Ensure the following ports are open in the Windows Firewall for inbound traffic:

Port Protocol Purpose
1433 TCP SQL Server default instance
5022 TCP Database mirroring / AG endpoint
1434 UDP SQL Browser (disable if not needed)

If a non-default port is used for SQL Server (recommended for security — see [[../Security/Security-Practices|Security Practices]]), substitute that port for 1433 and ensure the firewall rules match.

# Verify existing SQL-related firewall rules
Get-NetFirewallRule | Where-Object { $_.DisplayName -like '*SQL*' } |
    Get-NetFirewallPortFilter | Select-Object Protocol, LocalPort

Set TCP autotuning to normal:

netsh int tcp set global autotuninglevel=normal

# Verify
netsh int tcp show global

Set OS power plan to High Performance:

powercfg /setactive SCHEME_MIN

# Verify
powercfg /getactivescheme

After completing all configuration, proceed to [[SQL-Cluster-Installation|SQL Server Cluster Installation]].

See Also

  • [[Clustering|Clustering Index]]