September 18, 2012 at 5:08 am
Hi All
After going through some documentation on SQL Server 2012 HA, I need to make sure that my understanding is correct.
For a simple 2 Node Availibility group setup, the following would need to be in place
Two Servers, each with it's own seperate storage
Server1 and Server2 are joined to the same Windows Cluster
Both Servers have SQL installed as non clustered instances
Would this configuration allow for a 2012 HA group ?
How would I incorporate the High Availibility Failover instances with this? Would I still have seperate storage?
Thanks
September 18, 2012 at 5:55 am
Yes. If you're trying to set up availability groups, you are dealing with different servers with different storage. Now, that storage could be the same SAN, but each server is seeing it's own LUN or disk. They are not sharing a common set of database files. They're separate.
If it helps to think about it this way, it's how I explain. First, there was log shipping. We copy a log backup from one server to another and restore it. This was improved with mirroring. Similar situation, but more sophisticated operations. Now we have Availability Groups which are just advanced mirroring, or really advanced log shipping. The data & structures are communicated, but the files are not shared.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
September 18, 2012 at 5:59 am
Grant Fritchey (9/18/2012)
Yes. If you're trying to set up availability groups, you are dealing with different servers with different storage. Now, that storage could be the same SAN, but each server is seeing it's own LUN or disk. They are not sharing a common set of database files. They're separate.If it helps to think about it this way, it's how I explain. First, there was log shipping. We copy a log backup from one server to another and restore it. This was improved with mirroring. Similar situation, but more sophisticated operations. Now we have Availability Groups which are just advanced mirroring, or really advanced log shipping. The data & structures are communicated, but the files are not shared.
Thanks
How would I go about protecting my instance like I do with Failover Clustering? If I switch to HA groups?
Another thing, is this really a feasible way to go if all of your SQL instances only have 1 database each?
Thanks
September 18, 2012 at 6:06 am
SQLSACT (9/18/2012)
Grant Fritchey (9/18/2012)
Yes. If you're trying to set up availability groups, you are dealing with different servers with different storage. Now, that storage could be the same SAN, but each server is seeing it's own LUN or disk. They are not sharing a common set of database files. They're separate.If it helps to think about it this way, it's how I explain. First, there was log shipping. We copy a log backup from one server to another and restore it. This was improved with mirroring. Similar situation, but more sophisticated operations. Now we have Availability Groups which are just advanced mirroring, or really advanced log shipping. The data & structures are communicated, but the files are not shared.
Thanks
How would I go about protecting my instance like I do with Failover Clustering? If I switch to HA groups?
Another thing, is this really a feasible way to go if all of your SQL instances only have 1 database each?
Thanks
See my rather excellent articles at these links
http://www.sqlservercentral.com/articles/AlwaysOn/91408/[/url]
http://www.sqlservercentral.com/articles/FCI/92196/[/url]
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 18, 2012 at 6:16 am
I agree, read Perry's stuff.
With Availability Groups, you are getting something similar to a failover cluster, but without the stupid shared storage requirement (I always thought that was the silliest thing). If you have a listener associated with the AG, you can get automatic failover without having to do anything special to your app code (unlike mirroring). AG really is a huge improvement over the other HA functions.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
September 18, 2012 at 6:55 am
Grant Fritchey (9/18/2012)
I agree, read Perry's stuff.
Hey thanks Grant, i appreciate the vote of confidence
With AlwaysOn features it literally consolidates the functionality of mirroring and failover clustering into one neat package, although you can still combine FCI's with AO and that's what my 2nd link details 😉
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 20, 2012 at 3:05 am
Perry Whittle (9/18/2012)
Grant Fritchey (9/18/2012)
I agree, read Perry's stuff.Hey thanks Grant, i appreciate the vote of confidence
With AlwaysOn features it literally consolidates the functionality of mirroring and failover clustering into one neat package, although you can still combine FCI's with AO and that's what my 2nd link details 😉
Thanks
I've had a read-through the articles
The basic config of the AG is pretty simple, however, is it just me or does it become very complex when you start combining FCI ?
Thanks
September 20, 2012 at 3:53 am
SQLSACT (9/20/2012)
however, is it just me or does it become very complex when you start combining FCI ?Thanks
yes extremely, why do you think i wrote the article 😉
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 20, 2012 at 3:57 am
Perry Whittle (9/20/2012)
SQLSACT (9/20/2012)
however, is it just me or does it become very complex when you start combining FCI ?Thanks
yes extremely, why do you think i wrote the article 😉
Great stuff
Considering that this is where our company wants to go, I've got a lot of testing and research to do
September 27, 2012 at 4:25 am
Perry Whittle (9/20/2012)
SQLSACT (9/20/2012)
however, is it just me or does it become very complex when you start combining FCI ?Thanks
yes extremely, why do you think i wrote the article 😉
Hi Perry
I had a read through the articles, I am comfortable with the basic set up on an AG without incorporating FCI.
I just wanted to clear something up,
For a setup of an AG between 2 servers without incorporating FCI, I just need 2 servers, with it's own storage, right?
If I wanted a setup incorporating FCI with my AG, I'm gonna need 4 servers to accomplish this?
Thanks
September 27, 2012 at 5:30 am
SQLSACT (9/27/2012)
For a setup of an AG between 2 servers without incorporating FCI, I just need 2 servers, with it's own storage, right?
That is correct, recall that an availability replica cannot be based on the same host as its partner replica. Both servers must also be joined to the same Windows domain and cluster.
SQLSACT (9/27/2012)
If I wanted a setup incorporating FCI with my AG, I'm gonna need 4 servers to accomplish this?Thanks
depends on the scenario. 3 servers would service the following scenario
FCI called "SQLCLUST01\INST1" across Clusternode1 and Clusternode2.
Non clustered instance "Clusternode3\INST2" installed to Clusternode3.
AO group called "AOGrp1" created between instances "SQLCLUST01\INST1" and "Clusternode3\INST2"
My scenario in the article uses 3 SQL Server instances (1 clustered and 2 non clustered) and this would require 4 machines. This simulated a clustered prod instance and a DR and a maintenance instance.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 27, 2012 at 5:43 am
Perry Whittle (9/27/2012)
SQLSACT (9/27/2012)
For a setup of an AG between 2 servers without incorporating FCI, I just need 2 servers, with it's own storage, right?That is correct, recall that an availability replica cannot be based on the same host as its partner replica. Both servers must also be joined to the same Windows domain and cluster.
SQLSACT (9/27/2012)
If I wanted a setup incorporating FCI with my AG, I'm gonna need 4 servers to accomplish this?Thanks
depends on the scenario. 3 servers would service the following scenario
FCI called "SQLCLUST01\INST1" across Clusternode1 and Clusternode2.
Non clustered instance "Clusternode3\INST2" installed to Clusternode3.
AO group called "AOGrp1" created between instances "SQLCLUST01\INST1" and "Clusternode3\INST2"
My scenario in the article uses 3 SQL Server instances (1 clustered and 2 non clustered) and this would require 4 machines. This simulated a clustered prod instance and a DR and a maintenance instance.
Thanks, it's starting to click
Please confirm something
FCI called "SQLCLUST01\INST1" across Clusternode1 and Clusternode2.
Non clustered instance "Clusternode3\INST2" installed to Clusternode3.
AO group called "AOGrp1" created between instances "SQLCLUST01\INST1" and "Clusternode3\INST2"
The above setup is an AG combined with FCI,
Clusternode1 and Clusternode2 each have it's own storage (For Database) and they also have some shared LUNs attached from a SAN (For Windows and SQL dll's)
Am I on track here?
Thanks
September 27, 2012 at 6:09 am
SQLSACT (9/27/2012)
FCI called "SQLCLUST01\INST1" across Clusternode1 and Clusternode2.Non clustered instance "Clusternode3\INST2" installed to Clusternode3.
AO group called "AOGrp1" created between instances "SQLCLUST01\INST1" and "Clusternode3\INST2"
The above setup is an AG combined with FCI,
Yes it is.
SQLSACT (9/27/2012)
Clusternode1 and Clusternode2 each have it's own storage (For Database) and they also have some shared LUNs attached from a SAN (For Windows and SQL dll's)Am I on track here?
Thanks
No, the shared LUNs on these 2 servers are for the clustered SQL server instances data and log files, the local storage on each node will be the boot and system drives.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
September 27, 2012 at 6:35 am
Perry Whittle (9/27/2012)
SQLSACT (9/27/2012)
FCI called "SQLCLUST01\INST1" across Clusternode1 and Clusternode2.Non clustered instance "Clusternode3\INST2" installed to Clusternode3.
AO group called "AOGrp1" created between instances "SQLCLUST01\INST1" and "Clusternode3\INST2"
The above setup is an AG combined with FCI,
Yes it is.
SQLSACT (9/27/2012)
Clusternode1 and Clusternode2 each have it's own storage (For Database) and they also have some shared LUNs attached from a SAN (For Windows and SQL dll's)Am I on track here?
Thanks
No, the shared LUNs on these 2 servers are for the clustered SQL server instances data and log files, the local storage on each node will be the boot and system drives.
Something like this?(G Drive is the location on the data and log files)
Clusternode1 - G Drive(Shared)
Clusternode2 - G Drive(Shared)
Clusternode3 - G Drive(Local)
September 27, 2012 at 7:36 am
Have you worked with windows server failover clusters before, do you understand the concepts?
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
Viewing 15 posts - 1 through 15 (of 26 total)
You must be logged in to reply to this topic. Login to reply