As we have been going along we have been refining the two node Hyper-V Failover Cluster setup.
We are now at the point where the following base configuration would provide full Hyper-V failover capabilities for a very low cost:
- Hyper-V Server 2008 R2 SP1 Host (Requires 2)
- Intel Server System SR1695GPRX2AC
- Intel Xeon Processor X3470, 32GB ECC RAM, Dual 3Gbit SAS Controllers (SFF-3470 connectors).
- No local storage.
- Pair of SFF-3470 to SFF-8088 SAS Cables
- Promise VTrak E610sD Dual Controller RAID Storage System
- 10x 300GB 15K Seagate SAS (we use RAID 10 = 1.5TB usable)
- 2x 300GB 15K Seagate SAS global hot spares.
- Nodes boot from 75GB masked LUN on VTrak to eliminate the need for local storage.
- LUN Masking: Only the node will see its dedicated LUN as available.
With the above hardware configuration we are able to start our 2 node failover clusters at around $18K for the hardware setup.
For any client that currently has two or more servers and are conscious of the need for uptime we are promoting this setup.
At this price, there is no reason why we can’t be deploying Hyper-V failover clusters into most clients that have two good sized servers up for refresh ($6K-$8K/Server) or multiple small ones.
A big plus when talking about failover is the portability of the VMs. We can restore _anywhere_. Because of this portability we can promote our iSCSI backup target setup which then can eliminate the cost of any third party backup solutions.
Our goal is to go full on Windows OS native with as many tools as possible. And so far, we are being quite successful at making it happen.
Philip Elder
MPECS Inc.
Microsoft Small Business Specialists
Co-Author: SBS 2008 Blueprint Book
*Our original iMac was stolen (previous blog post). We now have a new MacBook Pro courtesy of Vlad Mazek, owner of OWN.
What happens when your VTrak goes down hard?
ReplyDeleteThe point I'm trying to make is this more expensive solution still has single points of failure in a more complex configuration.
ReplyDeleteWhat specific problem are you solving with a more expensive and more complex solution? Why wouldn't two nodes with local storage and a CDP system suffice?
Wondering how you manage the cluster? I have used a full installation of W2k8 R2 to manage local Hyper-V machines, but understand that using SCMVMM is preferred with Hyper-V server.
ReplyDeleteChris,
ReplyDeleteWe are recovering onto a node RPQ (Right Properly Quick)!
Yes, we are still in a position where there are risks. But, to mitigate those risks would require mirroring of the DAS/SAN solution.
The primary problem we are looking to address is immediate failover in the event of a server hardware failure which we _have_ experienced over the years.
We have been hard testing the VTrak units for about a year and a half now with little to no issues. So, as a storage platform it is an excellent start.
After discussing the CDP option with you via e-mail, it looks like a great idea. The big thing though will be what is lost via "failover" and whether that "failover" would be foolproof as well?
Albee,
We use the native Failover Cluster Manager, Hyper-V Management, and the Computer MMCs to manage our clusters.
SCVMM is _not_ a requirement to run a Hyper-V failover cluster.
Thanks,
Philip
Just to be clear - are you loading the baremetal Hyper-V Server 2008 R2 or just using the Hyper-V role in a full installation of Windows 2008 R2?
ReplyDelete