February 3, 2017 at 11:41 am
Hi,
We've 2 instances we want to cluster with FCI (without Always ON). There are 4 servers available to carry out this task.
It's better to build 2 FCI with 2 servers for each one, or 2 FCI with 4 servers? Any advantage ? I'm not sure what to choose.
February 5, 2017 at 3:03 pm
You give yourself far more flexibility by creating a single Windows Failover Cluster out of all four servers, and installing both FCIs on each of them. Are there any security restrictions in place that would prevent the data hosted in these instances from sharing a cluster host?
If you wish to leave yourself open to using Availability Groups (AG + FCI work together), place all four servers in a single cluster, but install FCI and zone storage on only half of them (install FCI1 on Server1 and Server2, install FCI2 on Server3 and Server4). They can replicate between each other, or you can add additional FCIs on those servers for AG replicas later.
Although it was written for older versions of SQL Server and Windows Server, the key concepts from this Microsoft whitepaper still stand: https://msdn.microsoft.com/en-us/library/jj215886.aspx
With Software Assurance, the SQL licensing on a 4-node, two-instance cluster is the same as building two separate 2-node clusters with one FCI each.
-Eddie
Eddie Wuerch
MCM: SQL
February 9, 2017 at 8:20 am
I agree on the use of all 4 nodes and putting 2 FCIs in there. It gives you multiple failover targets and no single point of failure on a 1:1 scenario. I would put them all in 1 WSFC and then 2 FCIs using all nodes as failover targets.
February 9, 2017 at 9:55 am
be careful with this, having a large number of cluster nodes with FCIs installed across all noeds will affect the installer and patch runtime.
On large clusters with lots of FCIs it can take a couple of hours or more just for the SP installer to reach the welcome screen as it carries out a full cluser enumeration each time it runs and on each node 😉
Install the FCIs only on selected nodes, if a further node is required later the add node wizard can be invoked
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
February 9, 2017 at 10:16 am
True, but 4 nodes would not be a large number in my opinion. I typically like having a max of 6 nodes and typically have 2 failover targets per instance so there is a mix.
Mileage may vary, but that is my opinion of setup.
February 9, 2017 at 10:34 am
dbaduck
It gives you multiple failover targets and no single point of failure on a 1:1 scenario.
FCIs do have a SPF, the shared storage!
dbaduck
I would put them all in 1 WSFC and then 2 FCIs using all nodes as failover targets.
this surely means you're adding every node as a failover partner which is the opposite of this below
dbaduck
I typically like having a max of 6 nodes and typically have 2 failover targets per instance so there is a mix.
this means you only install each FCI on 2 nodes of the cluster
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
February 9, 2017 at 9:58 pm
I understand Perry all you have said. I was simply illustrating that I don't like going over 6 nodes in each WSFC and I typically have any FCI having at least 2 failover targets exclusive of the one that is its home so technically there are 3 targets that it can sit on.
Using Shared Storage you certainly have an SPF, but I was not speaking of the storage :-).
February 10, 2017 at 3:56 am
ok thanks for clarifying
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply