Lambton
asked on
Correct steps to add a fourth node to windows cluster with CSV on fiber IBM SAN
Currently we have a 3 node Windows 2012 R2 Failover Hyper-V Cluster. (HVHOST1, HVHOST2, HVHOST3)
Each node has a Qlogic 2 port fiber card that each port is connected to an IBM V5000 Storage Controller.
On the V5000 there are 2 volumes mapped to these nodes, a witness volume and the CSV for the VM's.
Each node also has an iSCSI connection to another CSV on a Windows iSCSI Target Server.
Each node has 8 NIC's teamed
We use System Center VMM (SCVMM) to manage the Hyper-V cluster.
We need to add a fourth node to the cluster.
I've built a Windows 2012 R2 server with the same specs as the other 3 (same nics, qlogic cards, etc)
The server is a member of the domain.
The SAN has been configure to add the host, but does not have the volumes mapped yet (see screen shot - HV-EE-COLSAN.png)
MPIO has been installed (see screen shot - HV-EE-MPIO.png)
My question is, what is the best or only correct method to add this member to the existing hyper-v cluster. The existing cluster is in production, so please indicate if there will be downtime - I expect none, but just in case...
Thanks very much everyone.
HV-EE-COLSAN.png
HV-EE-MPIO.png
Each node has a Qlogic 2 port fiber card that each port is connected to an IBM V5000 Storage Controller.
On the V5000 there are 2 volumes mapped to these nodes, a witness volume and the CSV for the VM's.
Each node also has an iSCSI connection to another CSV on a Windows iSCSI Target Server.
Each node has 8 NIC's teamed
We use System Center VMM (SCVMM) to manage the Hyper-V cluster.
We need to add a fourth node to the cluster.
I've built a Windows 2012 R2 server with the same specs as the other 3 (same nics, qlogic cards, etc)
The server is a member of the domain.
The SAN has been configure to add the host, but does not have the volumes mapped yet (see screen shot - HV-EE-COLSAN.png)
MPIO has been installed (see screen shot - HV-EE-MPIO.png)
My question is, what is the best or only correct method to add this member to the existing hyper-v cluster. The existing cluster is in production, so please indicate if there will be downtime - I expect none, but just in case...
Thanks very much everyone.
HV-EE-COLSAN.png
HV-EE-MPIO.png
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
To correct, it is not my article, it is my reference to an article someone else wrote/created.
ASKER
Thanks Arnold, the article is good, however it doesn't have detailed steps regarding the FC San Configuration, which is where my knowledge lacks. @Philip Elder you mentioned to "make sure the Fibre Channel domains and LUN setup are identical", that's where I don't quite understand. The only LUNs I have created are CSV, so I'm attaching to the exact same LUN s that the other nodes are connected to, so I'd have thought that they are identical. FC Domains is not something that I'm familiar with and I don't see anywhere to configure that. I'll explain what I've done so far and maybe this will help put clarity to what I have/need to do.
I've setup windows server 2012 r2, installed MPIO, copied the MPIO supported devices from the other hosts. Connected to a test LUN on the SAN and it all seems to work, but I want to be dead sure before I start connecting to a production LUN. I don't see anywhere to configure FC Domains however, could you help me with that?
This is an IBM Storwize V5000 SAN btw with Qlogic HBA's.
Thanks,
Steve
I've setup windows server 2012 r2, installed MPIO, copied the MPIO supported devices from the other hosts. Connected to a test LUN on the SAN and it all seems to work, but I want to be dead sure before I start connecting to a production LUN. I don't see anywhere to configure FC Domains however, could you help me with that?
This is an IBM Storwize V5000 SAN btw with Qlogic HBA's.
Thanks,
Steve
THE FC is configure the zoning on the switches into which you plug the pair of feeds in the same way your current nodes are
This is a host for VMs in active/active meaning each host has VMs running on them..
i.e. do you have single or multi-path
node1 feed1 feed2
node2 feed1 feed2
node3 feed1 feed2
newnode feed1 feed2
the use of a new volume of 1GB deals with passing validation that a storage that is assigned to this host can be accessed by this host as the storage group assigned to the cluster can only be validated on the currently active node.
on the FC Swith/Switches you would need to drop feed1 of the newnode into the same zoning group as the other nodes, and the same on the node3.
Much depends on your topology.
This is a host for VMs in active/active meaning each host has VMs running on them..
i.e. do you have single or multi-path
node1 feed1 feed2
node2 feed1 feed2
node3 feed1 feed2
newnode feed1 feed2
the use of a new volume of 1GB deals with passing validation that a storage that is assigned to this host can be accessed by this host as the storage group assigned to the cluster can only be validated on the currently active node.
on the FC Swith/Switches you would need to drop feed1 of the newnode into the same zoning group as the other nodes, and the same on the node3.
Much depends on your topology.
ASKER
Oh OK I get it, it's a FC switch thing. That we don't use currently, we are directly attached to the SAN (4 hosts, 2 SAN controllers each with 4 FC ports), so each cluster node has 2 FC connections, one to each controller on the SAN.
/S
/S
Check stats on the qnap FC switch and whether it can sustain the throughput
Check whether FC switches added to the infrastructure will add LAG group each controller to switch ...
Check with qnap and whether the I/o demand you have is matched/met by the device.
Which FC hba's throughput do you have?
Check whether FC switches added to the infrastructure will add LAG group each controller to switch ...
Check with qnap and whether the I/o demand you have is matched/met by the device.
Which FC hba's throughput do you have?
To add to that, make sure the Fibre Channel domains and LUN setup are identical for the newly added node.
With that in hand, following the steps in @arnold's article should carry you through to an additional node.
Note that the word "identical" is absolutely critical.