October 9, 2012 at 1:13 am
I've recently bought some new servers and install SQL 2012.
I have 2 servers running on Windows 2008 Cluster with a shared SAN in one datacentre.
I have a backup SQL server in another datacentre.
I've added all 3 servers to the same cluster, which seems a little odd because the 3rd server cannot connect to the SAN so can never takeover but this is the only way I've found of having the 3 servers in the same cluster needed for the AlwaysOn HA stuff.
I've knocked the 3rd server's Vote to 0 so that when the interconnect between the 2 sites goes down and I shut down one of the primary servers the whole cluster doesn't shut down, this is all good.
However I need to be able to bring the 3rd server up if the 1st site is down but in its current configuration I cannot do this because there's no quorum. I don't want a failure of the second site to affect the cluster in anyway, i.e. if I start patching one of the primary's and the interconnect goes down I don't want the whole cluster to go down.
Is there anyway of doing this or do I have to revert back to using Mirroring between the 2 sites? I like the idea of failing over all the databases at the same time and having the virtual name move across, otherwise I have to run a script to failover all databases and then change DNS.
Anyone any thoughts?
October 9, 2012 at 5:06 am
darrenstokes (10/9/2012)
I've added all 3 servers to the same cluster, which seems a little odd because the 3rd server cannot connect to the SAN so can never takeover but this is the only way I've found of having the 3 servers in the same cluster needed for the AlwaysOn HA stuff.
Although it may seem odd, the 3rd server does not service the clustered instance and so, doesn't need to connect to the SAN. The machines are required to be part of the same cluster to be able to share the Cluster group that is created when creating an AlwaysOn Availability group
darrenstokes (10/9/2012)
I've knocked the 3rd server's Vote to 0 so that when the interconnect between the 2 sites goes down and I shut down one of the primary servers the whole cluster doesn't shut down, this is all good.
That sound fine, no problem here.
darrenstokes (10/9/2012)
However I need to be able to bring the 3rd server up if the 1st site is down but in its current configuration I cannot do this because there's no quorum.
No, in the event of disaster and total outage you would need to force quorum and bring the cluster service online.
darrenstokes (10/9/2012)
I don't want a failure of the second site to affect the cluster in anyway
This is why you removed the 3rd nodes vote. If you want the 3rd node to have a vote you will need to re think your number of votes within the cluster and their locations.
darrenstokes (10/9/2012)
Is there anyway of doing this or do I have to revert back to using Mirroring between the 2 sites? I like the idea of failing over all the databases at the same time and having the virtual name move across, otherwise I have to run a script to failover all databases and then change DNS.Anyone any thoughts?
have you created a listener for your AO group?
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
October 9, 2012 at 6:42 am
Thanks for the reply Perry.
That all ties in with what I've experienced so far.
What I don't like is that I can failover the AG to the 2nd site whilst the first site is still up and all is well until I shut down the network in the first site which I know I have to do soon, or if the interconnect goes down and then the cluster will shutdown. I know I can then force quorum but this isn't the desired effect. I think my only option is to go back to mirroring which I have already started. With mirroring I wont have any outage when the above quite likely scenarios occur.
Thanks again.
October 9, 2012 at 7:29 am
darrenstokes (10/9/2012)
With mirroring I wont have any outage when the above quite likely scenarios occur.
Depends on the mirroring mode you are using.
In any mode other than High Safety with auto failover you would have to Force service to bring the mirror online, pretty much the mirrroring equivalent of forcing quorum on the cluster node, although granted not as grand a task.
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply