How to eliminate the single point of failure in my cluster?

i've never configured cluster, i'm trying understand that's bugging me - how eliminate san single point of failure in cluster?

this caught interest: http://windowsitpro.com/windows-server-2012/windows-server-2012-building-two-node-failover-cluster

i'm trying conceptually understand how scenario work...

for maximum reliability, want cluster application runs on sql server. if understand correctly, i need @ least 2 servers , 1 san. increased reliability, should have 2 switches, , each node running sql have 2 connections - 1 each switch. far, pretty redundant...but san itself.

if want run given application in cluster , have fail on node, (in "perfect" world) need 2 sans well? if so, how cluster those? application automatically fail on live san?

thank you,

ed

i've never configured cluster, i'm trying understand that's bugging me - how eliminate san single point of failure in cluster?

this caught interest: http://windowsitpro.com/windows-server-2012/windows-server-2012-building-two-node-failover-cluster

i'm trying conceptually understand how scenario work...

for maximum reliability, want cluster application runs on sql server. if understand correctly, i need @ least 2 servers , 1 san. increased reliability, should have 2 switches, , each node running sql have 2 connections - 1 each switch. far, pretty redundant...but san itself.

if want run given application in cluster , have fail on node, (in "perfect" world) need 2 sans well? if so, how cluster those? application automatically fail on live san?

thank you,

ed

not easy. need not 2 sans sans doing synchronous mirroring between each other: data, caches , lock tables. or can go so-called dual controller solution cluster-in-a-box when 2 sans put single enclosure. that's hypervisor or os job figure out lun path dead , failover working 1 (mpio works here). application should not aware of underlying changes. that's way works. these days virtually san except real tier n hardware have ability create cluster between nodes or have dual controller option. there ways cluster w/o san: using nas (which easier manage, vmware esxi can nfs , hyper-v can smb), sas (something windows clusters had started long time ago, parallel scsi mean) , virtual san. virtual san (vsan) software taking multiple attached disks every server , creating distributed high performance san it. it's san w/o having 1 :) no hardware. esxi, kvm , xen can out of box , windows/hyper-v of third-party software. 

---

starwind iscsi san & nas

Windows Server  >  High Availability (Clustering)