1 / 17

Windows 2008 Failover Clustering Witness/Quorum Models

Windows 2008 Failover Clustering Witness/Quorum Models. Node Majority*. Recommended for clusters with odd number of nodes. Can sustain failures of half the nodes (rounding up) minus one Ex: a seven node cluster can sustain three node failures . No concurrently accessed disk required

adonia
Download Presentation

Windows 2008 Failover Clustering Witness/Quorum Models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Windows 2008 Failover Clustering Witness/Quorum Models

  2. Node Majority* Recommended for clusters with odd number of nodes • Can sustain failures of half the nodes (rounding up) minus one • Ex: a seven node cluster can sustain three node failures. • No concurrently accessed disk required • Disks/LUNs in Resource Groups are not concurrently accessed • Cluster stops if majority of nodes fail • Requires at least 3 nodes in cluster • Inappropriate for automatic failover in geographically distributed clusters Still deployed in environments that want humans to decide service location Cluster Status * Formerly “Majority Node Set”

  3. Node Majority with Witness Disk* Recommended for clusters with even number of nodes • Can sustain failures of half the nodes (rounding up) if the disk witness remains online • Ex: a six node cluster in which the disk witness is online could sustain three node failures • Can sustain failures of half the nodes (rounding up) minus one if the disk witness fails • Ex: a six node cluster with a failed disk witness could sustain two (3-1=2) node failures. • Witness disk is concurrently accessed by all nodes • Acts as tiebreaker Witness disk can fail without affecting cluster operations • Usually used in 2-node clusters, or some geographically dispersed clusters • Can work with SRDF/CE • 64 clusters/VMAX pair limit • Can work with VPLEX • Does not work with: • RecoverPoint • MirrorView Cluster Status * Formerly “quorum” disk

  4. Witness Disk Not generally recommended • Can sustain failures of all nodes except one • Loss of witness disk stops the cluster • Original “Legacy” cluster model for Windows until the introduction of Majority Node Set • Witness disk is the only voter in the cluster • Failure of the witness leads to failure of the cluster Cluster Status

  5. Node Majority with File Share Witness (FSW) Recommended for most Geographically Distributed Clusters • Can sustain failures of half the nodes (rounding up) if the FSW remains online • Ex: a six node cluster in which the disk witness is online could sustain three node failures • Can sustain failures of half the nodes (rounding up) minus one if the FSW fails • Ex: a six node cluster with a failed disk witness could sustain two (3-1=2) node failures. • Any CIFS share will work • FSW is not a member of the cluster • One host can serve multiple clusters as a witness • FSW placement is important • Third failure domain • Or FSW itself can be made to automatically fail over Timing issues can be a challenge • Works with no node limitations on: • SRDF/CE • MV/CE • RP/CE • VPLEX Cluster Status

  6. Why is Geographically Distributed Clustering a special case? • A two site configuration will always include a failure scenario that will result in majority loss • This is often desired behavior • Sometimes desirable to have humans control the failover between sites • Failover is automated, but not automatic • Simple to restart the services on surviving nodes (force quorum) • net start clussvc /fq • If automaticfailover between sites is required, deploy a FSW in a separate failure domain (third site)

  7. Things to note • Successful failover requires all disks in the resource group (RG) be available to the production node, including disks, requiring: • Replication between sites • A method to surface the replicated copies to the nodes in the DR site (Cluster Enabler), OR • A virtualization technique whereby the replicated LUNs are always available to the nodes in the DR site (VPLEX)

  8. Multi-site configurations

  9. Quorum Recommendations http://technet.microsoft.com/en-us/library/cc770830(WS.10).aspx#BKMK_multi_site

  10. FSW failure scenarios 3-site configuration – odd # voters Cluster Status

  11. FSW failure scenarios Even # voters Cluster Status

  12. FSW Failure Scenarios 2-site configuration – even # voters Cluster Status

  13. FSW Failure Scenarios 2-site configuration – odd # voters Cluster Status

  14. Node Weights • Nodes can be altered to have no vote • cluster . node <NodeName> /prop NodeWeight=0 • Useful when you have an unbalanced configuration • Hotfix required • http://support.microsoft.com/kb/2494036

  15. Dealing with loss of quorum • It’s an outage that requires manual intervention to recover • net start clussvc /fq • Cluster returns “unforced” when a majority of nodes come back online • Does not alter RPO of user data • Rolling failures may result in reversion of cluster configuration • FSW does not store cluster database (same behavior as disk witness)

  16. FSW Considerations • FSW can be SMB 1.0 • No need to be same OS as member nodes • Same domain, same forest • Cannot be a member of the cluster • Can be hosted on NAS (VNX) • 5 MB of free space • If Windows, server should be dedicated to FSW • 1 server can be witness to multiple clusters • Beware of dependencies • Administrator must have full control share and NTFS perms • No DFS

  17. Changing the quorum configuration Process for FSW • No downtime requirement • As long as there are a majority of nodes available • Account requirements • Administrators group on each node member • Domain user account • Create FSW • Start Failover Cluster Manager

More Related