10. Improved maintenance mode for disk resources. Need to take a snapshot of a disk or run chkdsk? Now the cluster will allow you to keep the resource online but unavailable for client use. What this means? No more downtime or reboots to do disk maintenance.
9. No more cluster HCL. We got rid of it. Now, the ball is in the customers court. Rather than the complicated process of having a vendor submit a solution as a whole out of cluster parts, the installation of cluster in 2008 now includes a validation test you run before you install the cluster feature. This test checks everything (storage, network, servers). As long as you have purchased hardware that's logo'ed for Windows 2008, and the validation test passes....you have a supported cluster.
8. Cluster nodes can now obtain their IP addresses from DHCP servers. There's no longer a requirement for clusters to have static IPs.
7. We now track cluster disks by 2 attributes, not just one. What this means...? No more disk failures due to signature changes. If a disk signature does change, we check another unique attribute of the disk to verify it's the correct disk and we "self heal". User never sees a thing. This is huge. Disk signature issues were easily one of our top supportability issues with NT4.0, 2000, and 2003.
6. No more quorum disk. We have moved away from the model where the quorum was a potential single point of failure. Our new model still uses a shared disk to keep the nodes registries in synch, but now, that disk also carries a "voting" mechanism. This disk, called the witness disk, gets one vote as does each node. As long as the cluster has enough votes (majority), the cluster stays online. This new model will tolerate a complete failure of the witness disk w/o impact to the user.
5. Cluster nodes can exist in different subnets. Now you can geo-cluster your servers in dispersed locations without the need to create an expensive and complicated VLAN over a WAN.
4. With VSS, you can now back up your cluster configuration settings. Say you accidentally delete your print spool resource and whack 1500 printers. With 2003 your recovery option was to manually recreate 1500 printers. Now, just run an authoritative restore of the cluster and return the settings to a previous point in time.
3. You can dynamically add disks to the cluster w/o any down time. Since by default, we leave any new drives presented to the OS in an offline state, with 2008, there's no risk of a volume becoming corrupted or having to shut down the entire cluster for that new disk to be recognized.
2. No more cluster service account. We use the 'local system' account to run the cluster service instead of a domain user. What that means, no more cluster service failures due to expired passwords, missing user rights, etc.
Microsoft, Windows Server 2008, Longhorn, Server, W2K8, Cluster, Clustering, Failover