Sie sind auf Seite 1von 16

1

This presentation gives an overview of Windows Server 2008 R2 Failover Cluster feature set. Failover clustering is a key piece of the datacenter as it provides high-availability to critical services, applications and VMs. We will not cover Hyper-V nor multi-site (disaster recovery) clustering as these are covered in other Microsoft Virtual Academy presentations.

Failover Clusterings goal in WS08 was all about simplicity, so now it is easy to use Failover Clusterings goals in WS08R2 was supporting Hyper-V, so now it has a great integration story

16-Nov-11 12:41

Part 1 will cover Hardware & Software as well as deployment Part 2 will cover Management, Supportability and HA Applications & Services

Understanding your software capabilities and licensing is an important planning consideration as Failover Clustering is a premium feature This means that it does not ship in WS Standard Edition, you need either Enterprise or Dataceneter Edition It is also available in Itanium (IA-64) edition, which is generally only used for SQL and it not commonly used, so we recommend deploying on x64 architecture. This SKU also supports Storage Server also contains Failover Clustering Hyper-V Server is a FREE version which we will cover in more detail on the next slide In WS08, Failover Clustering (WSFC) changed its name which was previously Microsoft Cluster Service (MSCS). This is important because when you are looking at documentation, you want to make sure that it refers to WSFC so you are looking at information for the current product as it was significantly architected between WS03 and WS08. The number of VM guest OS licenses is the biggest difference between the SKUs that support clustering. Hyper-V Server is FREE, but only licenses the host OS. Enterprise Server is generally cheaper than Datacenter, but it only contains 4 licenses for VM guest OS licenses (plus 1 for the host), while Datacenter contains an unlimited number of VM guest OS licenses (plus 1 for the host)

Hyper-V Server is completely free so download it today to try out Hyper-V and Failover Clustering! It contains all the Failover Clustering features, including CSV, Live Migration and 16-node cluster support. Since it is built on Server Core it is even more highly-available (more info on the next slide) However Server Core does not come with a GUI, but you can use PowerShell or CMD You can also connect to a Server Core system and manage it with a full GUI using Remote Server Administrator Tools (RSAT), also a free download and can be used from a client machine

Since Server Core does not contain utilities that are commonly exposed to the Internet, such as IE, Windows Media Player & Outlook Having fewer security vulnerabilities means fewer patches, which means fewer reboots and higher availabilty Server Core has no GUI it is recommended to use PowerShell However it can be managed using a full GUI through RSAT, which is a free download and even runs from Windows 7 Client machines

Have at least 2 networks for redundancy Networking with clustering is flexible use IPv4, IPv6 or 4to6 transition technologies Since 2008 it has supported DHCP which makes deployment and management easier, because you no longer have to worry about IP Addresses Nodes can reside on multiple subnets or a single subnet/VLAN, even if they are in different physical locations clustering supports this out of the box without any special configuration Internal cluster networks do more than just health checks they also share database updates and Cluster Shared Volumes (CSV) traffic NIC team recommended on every network for higher-availability if a NIC fails

Clustering is flexible in storage support and even added support for Fibre Channel over Ethernet (FCoE) The key requirement is that the storage must support Persistent Reservations, which is a SCSI-3 standardized command for allows one node to own a disk while preventing another one from stealing ownership of that disk. This is important for clustering as we need to enforce a single owner to avoid a split-brain scenario where partitioned nodes both try to own a disk which could potentially cause corruptions do not worry, this scenario is not possible with PRs. MPIO is recommended for high-availability and redundancy

Logo testing now includes cluster testing, so you should be confident that the hardware will work in the cluster Hardware does not even need to be identical although this is recommended so that the clustered workload behaves identically, regardless of which node it is hosted on. Cluster support is simple and significantly improved from the 2003 experience Encourage 2008 customers to move to 2008 R2 to gain significant feature improvements while still being able to use the same HW, including Cluster Shared Volume (CSV) support Clustering is supported by almost all OEMs as it is a critical component in the datacenter

10

11

Validate is the clustering Best Practice Analyzer It can be run before, during or after cluster deployment If the cluster has already been deployed, it can be used as a troubleshooting tool and it will also recommend best practices to give you the highest availability possible Microsoft Product Support will require you to run this before they assist you with troubleshooting it will save everyone time The test also inventories everything that is deployed in the cluster, which makes it easy for product support to understand how your cluster is configured, or for legal/compliance regulations All reports are automatically stored in the C:\Windows\Cluster\Reports folder for easy access at any time later

12

16-Nov-11

Cluster migration is the upgrade path You do not migrate a cluster. You migrate groups from one cluster to another cluster. It is not possible to simply upgrade a node and move it to the next version while it is still a member of the same cluster. This is because the cluster does not support nodes at different OS versions (2003, 2008, 2008 R2). Having nodes at different service pack levels (2008 R2 RTM, 2008 R2 SP1) is OK for the duration of the upgrade. For the first time in 2008 R2, all in-box clustered workloads have an upgrade path. Most use the Migration Wizard. The following have their own cluster-aware process (more info: http://technet.microsoft.com/en-us/library/ee791924(WS.10).aspx) Hyper-V: Import/Export or Backup/Restore Exchange: Move Mailbox SQL: Installation Wizard Upgrade DFS-R: Add the new cluster to the replication group and copy over settings Print: PrintBRM Migration is flexible, even with inconsistent configuration on the new cluster Enterprise Datacenter x86 x64 IA64 Static IP Dynamic IP IPv4 IPv6 Core Standard Physical Cluster Guest (Virtual) Cluster Different Subnets

13

Clustering is much easier and a critical piece of the datacenter / private cloud infrastructure Hardware support is flexible and easy it just has to pass Validate Validate is a great deployment, configuration and troubleshooting tool Migration enables upgrading clustered services to 2008 R2

14

15

16

Das könnte Ihnen auch gefallen