SQL Clusters on Nutanix and NOS 4.1.5

  • 20 September 2015
  • 11 replies
  • 9425 views

Userlevel 4
Badge +19
Hi All,

Just wanted to let you know that using traditional SQL failover clusters and SQL FCI, or any MSCS / Microsoft failover cluster is available in tech preview as of NOS 4.1.5. This is because we have introduced SCSI 3 Persistent Reservations to the iSCSI stack. To implement a cluster you need to use the MS iSCSI Initator and the Nutanix Volume Groups contract and iSCSI target for in-Guest iSCSI access. This will become a full GA feature in an upcoming release of NOS. Please experiment with it and let us know your thoughts.

11 replies

Userlevel 4
Badge +19
WIth 4.5 the failover clustering capability is now GA. If you wish to configure production SQL Clusters or other types of Windows Failover Clusters please upgrade to 4.5.
Badge +12
Is the best practice to have a 1:1 relationship of LUN:Volume Group or is it recommended/support to have a Multi:1 relaitonship

TempDBLUN : SQLVG
SysDBLUN : SQLVG
UserDBLUN : SQLVG

or
TempDBLUN : TempDBVG
SysDBLUN : SysDBVG
UserDBLUN : UserDBVG

Is there currently a process to grow a LUN with a VG?
(and) should I assume this is not a thin provisioned drive, therefore making growing the LUN even more important.

Currently on 4.5.0.2 and not seeing any information on resizing.

Will
Userlevel 4
Badge +19
Hi Will,

With 4.5.0.2 the sweet spot is 4 LUNs/vDisks per VG/Target. This is a performance sweet spot and not a limitation. You can configure more than that, but LUN's become CPU bound in the CVM with random IO and that will impact performance. 20K IOPS (8KB single IO pattern) or more is acheivable from a single LUN/Target, higher can be achieved with multiple LUN's/Target. Depending on the platform being used of course.

For a SQL DB in 4.5.0.2 you should be able to use a single Target/VG if you just need 4 drives. That would be the easiest configuration. Every LUN/vDisk is thin provisioned. We don't consume storage resources until the point that the application goes to write data.

At this stage I don't think we have a way to extend an existing vDisk/LUN connected to a target. So the process to expand the size would be to add another vDisk to the target and then use in guest tools to extend the volume. I'll check into this more and reply to this thread.
Userlevel 4
Badge +19
Hi 

Compression, Dedupe, and EC-X are all supported. As far as ADFS/NDFS is concerned the LUN is just a vDisk on our system just like any other VMDK would be.
Badge +3
Did you guys ever find out if the VG expand function is either possible or is on the roadmap?
Userlevel 4
Badge +18
Starting with 4.6 a VG disk can be expanded using acli (vg.disk_update), but the disks cannot be in use, which means the VG must be detached. Hot expansion is a known requirement, but not there today.
Badge +5
As an existing customer, the lack of online expansion of volume-group disks is preventing us from killing the SAN all together (windows failover clusters).

Do we expect this feature to come in the next or a near-future release? Specifically for Hyper-V customers.

-E
Userlevel 4
Badge +19
Hi Coleman,

This is already available and on the support portal and on the Nutanix bible. We have supported Windows Failover Clusters since AOS 4.1.5 and 4.5. There is also a best practice guide for Block Services availble on our web site. So it could be time to re-evaluate it.

http://www.nutanix.com/products/features/acropolis-block-services/

https://next.nutanix.com/t5/Nutanix-Connect-Blog/Putting-Acropolis-Block-Services-ABS-through-its-Paces/ba-p/11433

https://portal.nutanix.com/#/page/docs/details?targetId=Web_Console_Guide-Prism_v4_7:wc_block_services_requirements_r.html

https://portal.nutanix.com/#/page/docs/details?targetId=Command_Ref-Acr_v4_6:man_ncli_volume-group_auto_r.html

https://portal.nutanix.com/#/page/docs/details?targetId=Advanced_Admin-AOS_v4_7:vm__failover_cluster_create_t.html

https://vimeo.com/182407000

http://nutanixbible.com/#anchor-block-services-76

http://itbloodpressure.com/2015/10/09/nutanix-only-hci-vendor-with-windows-failover-clustering/

http://www.mcghem.com/2015/09/18/microsoft-sql-server-high-availability-options-on-nutanix/
Userlevel 4
Badge +19
Sorry, just read it's the online expansion of LUN's. That's a two phase problem. One Windows Clusters don't support online extention of LUN's unless you use Veritas. Secondly we will be supporting online expansion in our next software release. The way to expand until then is by adding an additional disk, which can be done online.
Badge +5
Thanks for the reply but I think you may have missed my concern.

Unless I'm mistaken, we are not able to resize disks that belong to a Volume Group without taking the Failover Cluster offline - that's what is preventing us from using NTX Block Services in our fialover clusters.

Am I mistaken and we can resize these disks while they're being used?

**Edit**
Just saw your second reply. Thank you. We are very much looking forward to that hot-expand as we're currently able to do this with our Luns that reside on our NetApp and Equilogic SANs.

Reply