So You Want to Build a Stretched Cluster?

Discusses some recommendations for building resilient stretched VMware vSphere clusters

Transcript

1. Before we start • Get involved! • If you use Twitter, feel free to tweet about this session (use hashtag #VMUG) • I encourage you to take photos or videos of today’s session and share them online • This presentation will be made available online after the event
2. So You Want to Build a Stretched Cluster? Some recommendations for building resilient stretched VMware vSphere clusters Scott Lowe, VCDX 39 vExpert, Author, Blogger, Geek http://blog.scottlowe.org / Twitter: @scott_lowe
3. Are you sure? • Stretched clusters are not a DR solution • No VM startup ordering • No orchestration of external resources • No inclusion of scripts or other tools • Stretched cluster requirements are somewhat steep • For many organizations, SRM with traditional replication is the right fit
4. • vSphere recommendations • Storage recommendations • Networking recommendations • Operational recommendations So you want to build a stretched cluster?
5. • Use vSphere 5 • Use DRS host affinity groups • Configure vSphere HA appropriately vSphere recommendations
6. • vSphere 5 eliminates some vSphere HA limitations • No more primary/secondary nodes • This leads to larger stretched clusters with more nodes in each site • Better handling of network isolation/network partition • More information on vSphere HA configuration shortly • vSphere 5 introduces the vMSC HCL category Use vSphere 5
7. • vMSC = vSphere Metro Storage Cluster • Introduces some new terms: • Uniform access = “stretched SAN” • Non-uniform access = “distributed virtual storage” • Provides boundaries for supportability of stretched cluster configurations New vMSC HCL category
8. • Allows you to mimic site awareness • Use PowerCLI to address manageability concerns • Use “should” rules instead of “must” rules Use vSphere DRS host affinity groups
9. • Managing DRS host affinity groups at scale does have some overhead; PowerCLI can help: • Use some sort of unique property to "group" VMs (could be a custom property if desired) • Use this grouping to automate VM placement into groups • Run the PowerCLI script regularly to ensure correct VM group assignment Using PowerCLI with host affinity groups
10. Configure vSphere HA appropriately • Configure per-site HA isolation addresses (using das.isolationaddress) • Set HA to run in N+2 failover capacity • Allows for host failures in each site • However, keep in mind that you can’t control the “+2” • Use per-site designated heartbeat datastores • Use “should” DRS host affinity rules to allow HA some flexibility
11. • Use storage from vMSC category • Be aware of storage performance considerations • Account for storage availability • Plan Storage DRS carefully • Use profile-driven storage Storage recommendations
12. • Consider cross-connect topology • Ensure multiple storage controllers at each site for availability • Provide redundant and independent inter-site storage connections • With VPLEX, use the third-site cluster witness Account for storage availability
13. • Align datastore clusters to site boundaries • Don't combine stretched/non-stretched datastores in a single datastore cluster • Understand the impact of SDRS on your storage solution Plan Storage DRS carefully
14. • Use user-defined capabilities to model site topology • You can use user-defined capabilities regardless of VASA support from the storage vendor • Create VM storage profiles to provide site affinity • This can help avoid operational concerns with VM placement Use profile-driven storage
15. • Plan for different traffic patterns due to VLAN extensions and L3 routing • Where possible, separate management traffic onto a vSwitch • Incorporate redundant and independent inter-site network connections • Minimize latency as much as possible Networking recommendations
16. • Account for backup/restore in your design • Handle inter-site vMotion carefully • Don't split multi-tier apps across sites Operational recommendations
17. • Consider a solution with client-side deduplication to reduce WAN traffic • A mechanism to reduce restore traffic would be nice to have as well • Might be able to leverage storage solution itself for restores • Restore to local side • Allow storage solution to replicate to remote side Backup/restore for stretched clusters
18. • An inter-site vMotion will impact DRS host affinity rules • An inter-site vMotion could require storage configuration updates • Review inter-site vMotions to: • Reconcile DRS host affinity rules and VM locations • Reconcile storage availability rules and VM locations • Impact on other operational areas Handling inter-site vMotion
19. Questions & Answers
20. Thank you!