November 15, 2011 at 8:17 am
I understand that this is an impossible question for anyone to answer perfectly without knowing how the system performs was simply hoping for some feedback re anything else i should be taking into account ref best practice , what other questions should i be answering.
In the process of designing storage with this point in mind:
-Protecting WRITE performance to SQL MDF/LDF
-MS Hyper-V hosting 2 virtual machines + space for 1 more.
-SQL 2008 standard (active passive cluster)
-Window Hyper-v cluster
-2 x servers hosting SQL cluster + windows cluster
-1 x SAN (2x fibre channel ctrls)
-14x600GB disks (1+hot spare)
-Hyper-v virtual machine used for data collection, important to isolate from SQL IO.
-Current SQL DB 600GB , provision for 3 x size of MDF growth taken into account.
-QUEST Litespeed compressed backup always less than third of MDF size
-Offsite backup provision in place for additional BAK files.
----------------------------------------------------------
Controller 0
ARRAY=Data
DISKS=8
RAID=10
GB=2400
VOLUMES= SQL Data 2235GB usable
ARRAY=Logs+TempDB+DTC+Quorum+BAK files
DISKS =4
RAID=10
GB=1200
VOLUMES=Logs+TempDB+Bak files 1111 usable
VOLUMES=DTC 5GB usable
VOLUMES=Quorum 1GB usable
Controller 1
ARRAY=Hyper-V
DISKS =2
RAID=1
GB=600
VOLUMES=Hyper-V 558GB usable to hyper V only
----------------------------------------------------------
Thanks for any pointers.
Scott
November 15, 2011 at 11:30 am
Not meaning to be rude, but to me this configuration makes absolutely no sense at all. As I understand it, you have a single HBA with two ports, where you use one port for Hyper-V and the other one for SQL Server. On top of all this, you have both a Hyper-V cluster and a SQL Server cluster of virtual machines.
Clustering is one of many high-availability options available. However, your system will never be more available than your weakest point. In this scenario, the disk system is only partially redundant. You have RAID in your SAN, but you use only one controller port for each LUN. So, if this port is going down, you're basically screwed. In a HA environment, each LUN should be present on two controllers, connected to different switches, to eliminate a single point of failure. Note, two controllers, not two ports. If a controller fails, it is likely that all ports are affected. If you want to separate Hyper-V and SQL Server traffic, you may however perfectly well use two dual port HBAs, where you use one port on each controller for Hyper-V and the two remaining ports for SQL Server.
Typically, you also use a similar configuration for your network.
An important principle for high availability: Never make any assumptions. For instance, if you assume that your HBA is working, it is a weak point. As an extention to the principle, we may also state: Eliminates any single point of failures.
November 16, 2011 at 12:49 am
Morning, not rude at all 🙂 , thank you for taking the time to reply.
I probably should have explained more clearly sorry. There are 2 x FC HBA cards. When i say controller 1 i meant HBA 1 as the "preferred path" , being that the other card would take up the job in the event of a problem with the other and vice versa. Sorry for the lack of clarity.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply