Professional Documents
Culture Documents
2012 - Fail-Over Clustering
2012 - Fail-Over Clustering
Fail-over Clustering
Geo Paul/ March 15, 2013
© Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Impact Areas
2 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Scalability & TCO
16 nodes 64 nodes
384 Virtual Workloads per node 1024 Virtual workloads per node
Enterprise & Data center Edition Standard & Data enter edition
3 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Tighter Integration with Hyper-V
•Infrastructure as a Service (IaaS) for Cloud scenarios (Private, Public, Hybrid
– Failover Clustering is considered the Infrastructure as a Service (IaaS) piece for Cloud scenarios (Private or Public)
•Virtual Machine failover prioritization
–priorities can be configured to control the order in which a virtual machine fails over or starts
•Enhanced Failover Placement
– Intelligence is built into the cluster service that determines the resources available on each candidate node before making placement
decisions
•Virtual Machine Mobility scenarios
– More than one simultaneous Live Migration
– Storage Migration and Hyper-V Replica
•Application Monitoring
– Solution to monitor applications running in a virtual machine
– Provide automatic corrective action such as restarting the VM, restarting a service within VM, or even migrating it to another node
• Guest Clustering improvements
– iSCSI-shared storage limitation removed
4
– Virtual Fibre Channel adapters (up to four virtual HBAs can be assigned to a single Guest)
© Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Patching Fail-over cluster
• Patching a cluster is an intensive, manual process requiring a good deal of planning and coordination
• Cluster-Aware Updating (CAU) feature provides inbox, automated process for patching a cluster
– Can be run outside of the cluster
– Can be run as part of the CAU Role implemented in a cluster
– Totally hands-off once a CAU run is started
• Works seamlessly with WU/MU or with WSUS
• If cluster is self-updating, it must have CAU role installed
• Foundational Technologies:
– Failover Clustering feature
– Windows Update Agent (WUA)
– Windows PowerShell 3.0
– Windows Management Instrumentation (WMIv2)
– Windows Remote Management (WinRM)
5 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Cluster-Aware Updating – Functional Description
• CAU role is installed using the CAU UI or the PowerShell cmdlet Add-CauClusterRole
• CAU role is not visible in Failover Cluster Manager. Run PowerShell cmdlet Get-CauClusterRole to
determine which node is the owning node
• Installed as part of the Failover Clustering feature
Dependencies
• Failover Cluster management tools
• .NET 4.x Framework installed
• Remote Management enabled on all cluster
nodes
• PowerShell Remoting enabled on each cluster
node (Enable-PSRemoting )
• Remote Shutdown firewall rule enabled on each
cluster node
7 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Active Directory Interoperability
8 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Active Directory Integration
9 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Domain-less cluster bootstrapping
• Legacy cluster required communication with a Domain Controller (PDC) before the cluster service could
start
• Virtualizing Domain Controllers in Hyper-V Failover Clusters were initially problematic as cluster could not
contact a DC to start the cluster service and bring resources online
• In WS2012, Failover Clusters no longer require DC connectivity to start. Uses local identity (CLIUSR) and
NTLM authentication
• NTLM using local account is fall back to Kerberos
– Local shared common identify is CLIUSR (local account) which uses NTLM
– Password stored in cluster database
– Password is rotated every 30 days
• Cluster Authorization Manager (CAM) is part of the cluster service architecture
– CAM uses local identity to control access to CSVv2 volumes
10 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Failover Clusters and Read-Only Domain Controllers
11 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Cluster Validation Process
12 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Failover Cluster Validation
• Series of tests that validates the hardware configuration to determine suitability for clustering
• Validation can be done before or after creating a cluster
– Recommended to do it before and after
• Cluster Configuration tests are not run before
• New cluster validation tests
– Hyper-V tests
• Only if Hyper-V role is installed on at least one node in the cluster
• Additional tests if Hyper-V Replica support is enabled
• Tests run against the virtual machines hosted in the cluster
• Failures in Hyper-V test does not mean cluster is not supported
– Storage test for Storage Spaces and Pools
– CVS tests included in Storage tests
– Can test a single disk using PowerShell (Test-Cluster –Disk), e.g: Test-Cluster -Disk "Cluster Disk 1"
Powershell :
Test-Cluster -Node node1.testlab.com,node2.testlab.com
13
Test-Cluster -Cluster testcluster
© Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Quorum Model Enhancements
•Dynamic Quorum model - new in Windows Server 2012
– Cluster manages the number of votes required to keep functioning
– Allows for cluster survival when >50% of the nodes are down
– Enabled by default
– Works with all Quorum models except Disk Only
•Implementation of Node Vote Weight
– Controls which nodes have a vote in determining quorum
– Useful in multi-site cluster scenarios
– Works with all Quorum models except Disk Only
14 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Dynamic Quorum model
• Implemented to provide higher availability within a cluster by continuously monitoring the quorum
model and making adjustments based on the active cluster membership
• This model allows a cluster to continue to function when less than 50% of the nodes are active
• Dynamic Quorum model is enabled by default when a cluster is created
• Using the Get-Cluster PowerShell cmdlet, the property DynamicQuorum is set = 1
• Dynamic Quorum can be manipulated using PowerShell or the Configure Cluster Quorum Wizard
• When a node leaves the cluster (Node is shut down), the node adjusts its own weight as part of the
shutdown of the cluster service and quorum is re-calculated by the remaining nodes
15 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Configuring Quorum Model in Windows Server 2012
• Use Configure Cluster Quorum Wizard in Failover Cluster Manager
– Use Typical Settings (Default and Recommended)
• Uses logic implemented in Windows Server 2008 (+R2)
– Add or Change Quorum Witness
• Allows for manual selection of a witness resource
– Configure Disk Witness (even # of nodes)
– Configure File Share Witness (multi-site clusters)
– Configure No Witness (odd # of nodes)
– Advanced Quorum Configuration and Witness Selection
• Configure Node Weights
– All Nodes
– Select Nodes
– No Nodes (witness disk is required and cannot fail)
• Use Set-ClusterQuorum PowerShell cmdlet to set the quorum model
• Use Get-ClusterQuorum PowerShell cmdlet to list the quorum details
16 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Node weighting
• The concept of node weighting is primarily implemented in a multi-site cluster scenario
• Node weights used to control the vote calculation process.
• Nodes are assigned a weight=0 if they will not be included in the quorum calculation by Quorum Manager
• Typical configuration is nodes at the Primary site have votes but nodes at DR site may not
– Determining factor is whether or not automatic or manual failover will be executed
• Does not work with “Disk Only Quorum model”
• Cluster Node Weights can be assigned using the Configure Cluster Quorum Wizard and by selecting the
option for Advanced Quorum Configuration and Witness Selection
17 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
HA Roles
• Cluster-Aware Updating (CAU) – Not listed in “Configure Role
section”
• DFS Name Server
• DHCP Server
• Distributed Transaction Coordinator (DTC) Deprecated Roles:
• File Server (File Server Role)
Highly Available Print Server (Spooler)
– Scale out file server
Remote Desktop Connection Broker Roles
– Generic Application\Script\Service
• Hyper-V Replica Broker (Hyper-V Role)
• iSCSI Target Server (part of File Server Role)
• iSNS Server (iSNS Server Feature)
• Message Queuing (Message Queuing Services Feature)
• Other Server
• Virtual Machine (Hyper-V role)
• WINS Server (WINS Server Feature)
18 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
• Task Scheduler (created in a hidden group, not visible in FCM)
Cluster Storage
• Cluster Shared Volumes version 2 (CSVv2)
– High speed CSV I/O redirection
• Integration with SMB multi-channel and SMB Direct resulting in zero downtime for connected client s
– Multi-subnet support (e.g. Multi-site cluster configurations)
– CSV diagnostic Logs (Application and Services Logs\Microsoft\Windows\)
• FailoverClustering-CsvFlt
• FailoverClustering-CsvFs
• Improved Persistent Reservation algorithm which is less demanding on storage arrays
– PR keys are now unique to each node
• Increased logging to the cluster log during physical disk online process
• Integration with Storage Spaces feature (Storage pool)
– Provide highly available Simple and Mirrored spaces (Parity not supported)
• iSCSI Software Target integration
• Maximizing file system availability - Failover Clusters take advantage of the new file system check and
repair functionality (CHKDSK)
19 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Cluster Shared Volumes (CSV) Version 2.0
• CSV first introduced in Windows Server 2008 R2
• CSV provides the following:
Distributed file access
Concurrent access to the CSV volume from any node in the cluster
Physical location of the data is transparent to the application
In case of Hyper-V, multiple VMs can share a common CSV volume
Live Migration of VMs is faster because storage ownership does not have to change
• CSV is compatible with any block level storage attached using FC, iSCSI, or SAS
• CSV requires the NTFS file system residing on either an MBR or GPT disk volume
• CSV file system CsvFs (CsvFs.sys) is mounted on top of NTFS volumes
• CSV in WS2012 supports arbitrary workloads as evidenced by supporting SMB shares in the Scale-Out File
Services Role
– CSV takes advantage of SMBv3 features like multi-channel and Direct Access (RDMA)
• Right-click on a disk to add it to the CSV Namespace
• Get-ClusterSharedVolume PowerShell cmdlet is ued to get CSV information
20 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
CSV requirements
• Client for Microsoft Networks and File and Printer Sharing protocols enabled
• NTFS formatted partitions are required. ReFS is not supported in this release
• NTLM authentication with a Domain Controller is no longer required. NTLM authentication between cluster
nodes is used
• Cluster validation process verifies basic requirements like protocol bindings and SMB share access to the
CSV default share
21 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
WS2012 CSV Improvements
22 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Cluster Networking
23 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Resource Hosting Subsystem (RHS)
25 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Scale-Out File Services
• Takes advantage of :
– New client access method by way of a new cluster resource type (Distributed Network Name (DNN))
– Cluster Shared Volumes v2 (CSVv2)
– SMBv3.0 . Cient SMB connections being distributed across all nodes in the cluster
• Supports SMB and NFS shares
• Provides for Continuous Availability using Transparent Failover.
• Hyper-V and SQL Server takes benefit of this feature
26 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Installing Failover Cluster Feature
• Deprecated functionality
– Failover Cluster Automation Server
– Failover Cluster Command Interface (cluster.exe CLI)
27 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Creating a Failover Cluster
Powershell:
• New-Cluster -Name testcluster -Node node1.testlab.com,node2.testlab.com -StaticAddress 192.168.20.150
• If DHCP is used and you do not want all storage added by default:
New-Cluster -Name testcluster -Node node1.testlab.com,node2.testlab.com -IgnoreNetwork -NoStorage
28 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Failover Cluster Logging and Reporting
29 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Failover Cluster Log
30 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Powershell commands
33 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
Thank you
© Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.