Failover Clustering: What's new in Windows Server 2012 R2John MarlinSenior Support Escalation EngineerMicrosoft

DCIM-B354

Session Objectives and TakeawaysOverview new feature enhancements coming in Windows Server 2012 R2 for Failover Clustering and availabilityRoadmap of incremental changes from Windows Server 2012

File Based StorageBlock Storage

Guest Clustering with Shared Virtual Disks

VHDX VHDX

Guest Clustering

Guest Clustering with commodity storage

Sharing VHDX files provides shared storage for Hyper-V Failover Clustering

Maintains separation between infrastructure and tenants

Virtual SAS

VM presented a shared virtual SAS disk

Appears as shared SAS disk to VM

Used for data disk only

Cluster Shared Volumes (CSV) on block storage

Scale-Out File Server for file based storage

Guest ClusteringGuest Clustering shared storage deployment options with Windows Server 2012 R2:

Windows Server 2012

Windows Server 2012

R2Fibre Channel P PiSCSI P PFile (SMB) P PShared VHDX P

SCVMM Hyper-V VM Settings

Creating Shared VHDX Example

Example of creating and attaching a shared VHDX to two existing VMs

PS C:\> New-VHD -Path C:\ClusterStorage\Volume1\Shared.VHDX -Fixed -SizeBytes 30GB

PS C:\> Add-VMHardDiskDrive -VMName Node1 -Path C:\ClusterStorage\Volume1\Shared.VHDX -ShareVirtualDisk

PS C:\> Add-VMHardDiskDrive -VMName Node2 -Path C:\ClusterStorage\Volume1\Shared.VHDX -ShareVirtualDisk

Shared VHDX

Technet

VM Drain on ShutdownVMs live migrated to another node during shutdownEnsures reboot / shutdown does not incur downtime to VMs for unknowing adminVMs moved to “Best Available Node”

Most free memory

Honors VM prioritizationEnabled/Disabled via the DrainOnShutdown cluster common property

Still recommended to drain nodes before shutdownRecommend

ed Technet

VM Component Health Detection

Live migration

Component level VM network health detection

Media sense loss detection of network disconnects

Health Monitoring

Live migrates VMs off hosts with no external network connectivity for the VM

Verifies destination node has the network available

Configurable per networkRecovery

Blog

Scale-out File Server

File Server 2

Optimized CSV placement policiesOptimized failover placement policies to spread CSV ownership across the clusterEnsures CSV volume ownership is distributed across the cluster

Placement Scenarios:CSV FailoverNode rejoining the clusterNew node added to the clusterCold start of cluster

File Server 1

Share2 Share1 Share2

SMB Client

\\SOFS\Share1 \\SOFS\Share2

Share1

Technet

Increased CSV ResiliencyIsolated SMB Instances

Separation for improved resiliencyIndependent processes and queues for regular SMB client traffic and inter-node SMB trafficImproves scale of inter-node SMB traffic between CSV nodes

Multiple SMB Server instances per Node

Default Instance – Handles incoming traffic from SMB clients accessing regular file sharesCSV Instance – Handles only inter-node CSV traffic (metadata access or redirected traffic)

Node 1

SMB Client

SMB Server Default Instance

SMB Server CSV Instance

SMB Client

Node 2

SMB Server Default Instance

SMB Server CSV Instance

SMB Client

CSV Dependency Health DetectionCluster Shared Volumes (CSV) health monitoring of the Server Service

CSV leverages SMB as a transport protocol for orchestrating and forwarding I/O between nodes

If the Server Service becomes unhealthy on a node, CSV will designate a new coordinator node

Cluster Shared Volumes

Server Service

Technet

Cluster Shared Volumes (CSV) Cache CSV cache supports allocating larger amounts of memory (up to

80% of total memory)

Most interesting for Scale-out File ServersScalability

Physical Disk resource EnableBlockCache enabled by default

Still need to enable via BlockCacheSize to allocate memoryDefaults

Recommended to enable CSV Cache for all Hyper-V and Scale-out File Server deployments

Greater allocation for Scale-out File Server as they are not memory bound

Recommended

Blog

Diagnosability

Improved logging when creating a cluster and adding a node

Enables pinpointing root cause quickly and easilyDeployment Logging

Increased logging to Operational channel for cluster resource state changes

For example: When cluster resources move from one node to another

State Logging

Cluster Shared Volumes Diagnosability

Displays on a per-node basis the CSV I/O mode and the reason if in a redirected mode

Direct I/O, Block level redirected mode, File system level redirected mode

Get-ClusterSharedVolumeState PowerShell cmdletCSV State

Diagnosability

ReFS with CSV

Cluster Shared Volumes Interoperability

Interoperability

Tiered Spaces

with CSV

Parity Spaces

with CSV

Dedup with CSV

Spaces Write-Back

Caching with CSV

Technet

Node Health Detection over NetworkCluster Health monitoring:

Nodes exchange heartbeats every 1 second (configurable)Nodes considered down if it does not respond to 5 heartbeats (configurable)

Default settings are fairly aggressive to deliver the highest levels of availability

However are fully configurable to meet your business needs

For a Hyper-V deployment slightly more relaxed settings may make sense

Traditionally the definition of down, is when clients cannot connect to an app in the VMIn general, TCP defines recoverable network errors for applications

Recommended for cluster heartbeats not to exceed 20 seconds

Greater resiliency to transient network failures with Windows Server 2012 R2

Heartbeat thresholds increased by default for Hyper-V ClustersCluster heartbeating improved for increased resiliency to packet loss

Cluster Property Default

Hyper-V Default

SameSubnetThreshold 5 10

CrossSubnetThreshold 5 20

BlogTechEd 2013

Cluster

Cluster Global UpdatesFor state change or database update nodes must be notified and updated

Leverages clusters Global Update Manager (GUM)

Two GUM modes now in Windows Server 2012 R2

Update

Greater than 50% of nodes must acknowledge before considered committed

Toggled on by default for Hyper-V clusters

Enables the cluster to move forward if there is a node with latencyMajorityAll nodes must acknowledge before considered committed

Default mode for all other workloads (SQL, Exchange, etc…)

Configurable with DatabaseReadWriteMode cluster common propertyAll

Technet

Network Validation ImprovementsNetwork validation improved from simple PING to using NetFT

Verifies port 3343 and full cluster network connectivity requirements

Provides better diagnosability and pre-identifies cluster configuration problems

Cluster DashboardNew Cluster Dashboard for Status at a GlanceFocused at multi-cluster management

Dynamic WitnessWitness vote dynamically/automatically adjusted based on cluster membership with dynamic quorum

Odd node votes (3) + no witness vote (0) = 3Even node votes (2) + witness vote (1) = 3

Automatic functionality based on WitnessDynamicWeight property.

Always configure a witness with Windows Server 2012 R2

Clustering will determine when it is best to use the Witness

Configure Disk Witness if shared storage, otherwise FSW

New Recommendati

on

Disk Witness1 / 0

Node1 / 1

Node1 / 1

Node1 / 1

Dynamic Witness

Validate

Failover Cluster Manager

Intuitive Quorum Configuration UINode vote weights and dynamic quorum status easy & quick to view

Removed legacy concepts of ‘quorum modes’

Node MajorityNode and Disk MajorityNode and File Share Witness Majority

Cluster

Site1

Tie BreakerCluster will survive simultaneous 50% loss of votes

Balanced multi-site clusters with complete site partition

One site automatically elected to winSite without LowerQuorumPriorityNodeID cluster common property winsNodes in the other site drop out of the cluster

Site2

Cluster

Force Quorum ResiliencyCluster detects partitions after a manual ForceQuorumForceQuorum’d partition is deemed authoritativePartitioned nodes restarted and re-joined

Cluster brought back into a single view of membership

Site1 Site2Manual Override

with ForceQuorum

Nodes Restarted

When Site2 partition detected

Reducing Cluster DependenciesActive Directory-detached cluster Network Names

Enables creating a cluster without computer objectsNew-Cluster –AdministrativeAccessPoint DNS

Simplifies cluster deploymentsBest fit for SQL Server Clusters

Flexibility to create clusters with or without Active Directory integration

Still required that Nodes are domain joined

Cluster

Active Directory

CNOVCO

Technet

Considerations with AD-detached ClustersWhat to consider when choosing a model

No computer objects, means no Kerberos authentication to the nameNTLM only client authentication against cluster names

Intra-cluster authentication still leverages Kerberos

Role Position Notes

MSMQ Clusters Not Supported MSMQ stores properties in AD

File Server Clusters Not Recommended Kerberos for SMB preferred

Hyper-V Clusters Not Recommended No live migration support

SQL Server Clusters Supported Best fit if using SQL Authentication

SQL Server 2014 Failover Cluster Instance now supported on Windows Server 2012 R2 Cluster Shared Volumes (DCIM-B364)

Simplified storage management with consolidation of multiple instances onto a single LUNBetter capacity planning and storage utilizationAddresses drive letter limitations allowing more than 24 SQL instances on a single cluster

SQL Server 2014 on Cluster Shared Volumes

Consolidation

Databases can be moved without requiring any drive ownership changesFaster Failover times with no dismounting and remounting of volumes requiredIncreased resiliency to storage faults with CSV I/O redirectionChkdsk integration for file system correction with zero downtime

Availability

CSV block level read-only cache for unbuffered I/O to SQL databases

Performance

SQL Server Availability Groups are now supported on an Azure VM

http://msdn.microsoft.com/en-us/library/jj870962.aspx

Azure has no shared storage, so clustering requires data replication

Failover Clusters on Windows Azure

SQL AGPrimary

SQL AGSecondar

y

Azure

SQL AGPrimary

SQL AGSecondar

y

Azure

Private

Public Cloud Hybrid Cloud

SiteC

SiteB

30s, 5m, or 15m replication frequencies

Variable Recovery Point Objective (RPO) for metro vs. geographically dispersed

Hyper-V Replica Improvements

SiteA

Configurable Replication

Frequencies

Near site and offsite Replication

Second hop can be equal to or greater replication frequencies than first hop

Multiple Replica’s

Windows Azure Hyper-V Recovery Manager (HRM) to manage datacenter level failover

Site-to-site failover management at scaleManagement

Hyp

er-V

Reco

very

Man

ag

er

Technet

ApplicationHA for Microsoft Hyper-V

Application Restart

Virtual Machine Restart

Virtual Machine Failover

Virtual Machine Migrate

• Recovery or Proactive based remediation

Recovery – app restart, VM restart, failover, image recovery based on failuresProactive – migrate based on trends toward service loss; workload close to exceeding capacity

• Customizable recovery behavior

App or VM restart limitIn-guest remediation only - Enable/Disable App HeartbeatTurn off remediation during planned maintenance - Enter/Exit Maintenance modeDownload the whitepaper at

http://www.symantec.com/connect/articles/download-new-whitepaper-applicationha-61-hyper-v

In Review: Session Objectives And Takeaways

Flexibility

Reliability

Interoperability

NIC health detection and VM failover

Dynamic Witness and survival of simultaneous 50% node loss

SMB isolated instances and Server Service health detection

CSV compatibility with Tiered Spaces, Parity Spaces, and Dedup

Guest Clusters with shared VHDX files

AD agnostic clusters without associated computer objects

Windows IT Pro Technet

Breakout SessionsDCIM-IL308 Windows Server 2012 R2: Introduction to Failover Clustering with Hyper-V (repeats on May 13 at 1:30 pm and May 13 at 5 pm)

DCIM-IL308-R Windows Server 2012 R2: Introduction to Failover Clustering with Hyper-V (repeated from May 12 at 3pm)

DCIM-IL308-RR Windows Server 2012 R2: Introduction to Failover Clustering with Hyper-V (repeated from May 12 at 3pm)

DCIM-H321 Windows Server 2012 R2: Introduction to Failover Clustering with Hyper-V

DCIM-B327 Cluster-in-a-Box Using Windows Server 2012 R2: Solutions and Applications

DCIM-B337 File Server Networking for a Private Cloud Storage Infrastructure in Windows Server 2012 R2

DCIM-B349 Software-Defined Storage in Windows Server 2012 R2 and Microsoft System Center 2012 R2

DCIM-B364 Step-by-step to Deploying Microsoft SQL Server 2014 with Cluster Shared Volumes

Related content

LabsDCIM-IL200 Build Your Storage Infrastructure with Windows Server 2012 R2 (repeats on May 14 at 5 pm)

DCIM-IL200-R Build Your Storage Infrastructure with Windows Server 2012 R2 (repeated from May 14 at 8:30 am)

DCIM-H205 Build Your Storage Infrastructure with Windows Server 2012 R2

Come Visit Us in the Microsoft Solutions Experience!

Look for Datacenter and Infrastructure ManagementTechExpo Level 1 Hall CD

For More InformationWindows Server 2012 R2http://technet.microsoft.com/en-US/evalcenter/dn205286

Windows Server

Microsoft Azure

Microsoft Azurehttp://azure.microsoft.com/en-us/

System Center

System Center 2012 R2http://technet.microsoft.com/en-US/evalcenter/dn205295

Azure PackAzure Packhttp://www.microsoft.com/en-us/server-cloud/products/windows-azure-pack

Resources

Learning

Microsoft Certification & Training Resources

www.microsoft.com/learning

msdn

Resources for Developers

http://microsoft.com/msdn

TechNet

Resources for IT Professionals

http://microsoft.com/technet

Sessions on Demand

http://channel9.msdn.com/Events/TechEd

Complete an evaluation and enter to win!

Evaluate this session

Scan this QR code to evaluate this session.

© 2014 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.