December 13, 2011 at 4:42 pm
We have a cluster where 2 nodes are running different DBs. When failing over one node (with all its clustered resources) switches to the other physical server.
Lets say If a disk fails on one server, the entire instance moves to the other server in this configuration. In case of a failover, once the second server is running both instances, will the resources fail back to the first server if a resource fails on the second server?
Thanks
December 13, 2011 at 9:45 pm
from your question I assume you are running an Active/Active Cluster?
when a resource in N1 fails, the whole resource group will fail over to N2, and if a resource fails on N2 WHILE N1 is available, it will failover to N1, if N1 is not available, your cluster will go down.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
This thing is addressing problems that dont exist. Its solution-ism at its worst. We are dumbing down machines that are inherently superior. - Gilfoyle
December 13, 2011 at 10:02 pm
Henrico Bekker (12/13/2011)
from your question I assume you are running an Active/Active Cluster?when a resource in N1 fails, the whole resource group will fail over to N2, and if a resource fails on N2 WHILE N1 is available, it will failover to N1, if N1 is not available, your cluster will go down.
Thanks. Yes, active/active. I assume if "Prevent failback" is checked in a situation when N1's resource fails and it fails over to N2 and then that same resource fails again, the node goes down, right?
December 13, 2011 at 10:57 pm
Thanks. Yes, active/active. I assume if "Prevent failback" is checked in a situation when N1's resource fails and it fails over to N2 and then that same resource fails again, the node goes down, right?
The node itself wont go down, but the Cluster group wont function, but it also depends on what exactly fails...
if its a motherboard, local disk running the OS, then yes, whole server could go down...but if only a dependancy for a resource group goes down, chances are the server will still run.
Node is just another way of saying Server x in the Cluster...
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
This thing is addressing problems that dont exist. Its solution-ism at its worst. We are dumbing down machines that are inherently superior. - Gilfoyle
December 14, 2011 at 12:46 am
Lexa (12/13/2011)
We have a cluster where 2 nodes are running different DBs. When failing over one node (with all its clustered resources) switches to the other physical server.Lets say If a disk fails on one server, the entire instance moves to the other server in this configuration. In case of a failover, once the second server is running both instances, will the resources fail back to the first server if a resource fails on the second server?
Thanks
Definitely. That is the whole purpose of clustering, to provide the high availability to SQL Server.
December 14, 2011 at 12:56 am
As said, yes: Resources may fail back and forth over time, and that is the whole purpose of a cluster. But, you barely touch an important area here. How about disks which is not required for keeping your services up and running? You may for instance have backup disk, which SQL Server backs up its databases to. If this disk should fail, the affected instance will start to fail back and forth between the cluster nodes. This is in my opinion not a good option, the backup disks are not required for keeping the database up and running. You definitely want to monitor this disk to get an alert, but I would recommend turning "affect cluster group" off for this disk. If it fails, it would most likely fail on the other node as well.
December 14, 2011 at 1:46 am
You highlighted a good & valid point here.:cool:
Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply