Whatcha talkin’ bout? FOO! Setting up proper fail over in a Cluster.

Here is what a 2 node fail over cluster should look like. Double network redundancy on the back end with each node and the SAN connecting to both back end switches.

In order to setup proper cluster fail over the Server needs to be set to Fail Over Only (FOO). Remember that Windows has iSCSI volume size restrictions. So when you create volumes and LUNs on you SAN you need to limit the size. See link at end of article.

To do this you need a couple things, first you need to connect the iSCSI connection to both servers. Some SAN manufacturers have their own DMS drivers, which are usually a modified version of the Microsoft DSM driver. HP actually recommends using the Microsoft DSM.

The proper DSM is required in order to setup proper MPIO (Multipath Input and Output) on a cluster.

In short MPIO, is the multipath interconnect necessary for failover, it uses the DSM driver to achieve this. Generally the DSM driver is provided by the OS vendor, in this case Microsoft. This is also the HP recommended method of connecting to the SAN from Microsoft Windows, and most other manufacturers also use the Microsoft DSM driver.

Map iSCSI connections

First we need to properly map the iSCSI connections. Be aware that you will be mapping the same connection multiple times, this is necessary for failover. In the above example each server has 4 connections. 2 for the 20 subnet and 2 for the 30 subnet. Open up iSCSI Initiator and select the Discovery tab.

In the discovery tab add all 4 IP destinations, x.x.20.110, x.x.30.110, x.x.20.111, x.x.30.111.

Click the Discover Portal… button and add each one of those connections.

Next select the Targets tab, you should see the inactive iSCSI connection here.

Highlight the connection and click Connect. The HP SAN is setup with a single iSCSI connector and multiple LUNs. Some devices have multiple iSCSI connectors with a single LUN on each. Depending on the setup you might have to do this to each connector.

A Connect To Target window will pop up, check off Enable multi-path and click Advanced.

Under Local adapter select Microsoft iSCSI Initiator, for the Initiator IP select the IP for the Server, and the Target Portal IP should be one of the two IPs on the same subnet as the Initiator IP. It should look like the following.

Now repeat these steps, highlight the same connection, click Connect, check of Enable multi-path, click Advanced…, rinse and repeat, this will map the other three connections.

10.10.20.4 -> 10.10.20.111, 10.10.30.6 ->  10.10.30.110, 10.10.30.6 -> 10.10.30.111.

If you click the Favourite targets tab you should see 4 similar targets. These are all the connections you just created for the one iSCSI target(iqn).

Set up Connection Fail Over

Next start the MPIO applet, Start > Run > mpiocpl your Vendor should be listed in the Devices: window. If it is not you will need to add it via the Discover Multi-Paths tab, others window. Highlight the Device Hardware and click Add. Say no to the reboot.  

Next in the SPC-3 compliant window check off Add support for iSCSI devices and click Add. You will again be prompted to reboot. This time do so.

If you run the command mpclaim -s -d in an admin CMD session you should see the connection now.

Back in the iSCSC Initiator applet, if you highlight the iqn connection on the Targets tab and click Properties, Devices, and MPIO, you should see the Load Balance policy and all the paths that this connection can fail over to.

Your load balance policy will initially default to Round Robin change this to Fail Over Only. If you do this all but one connection should set to Active, all others will go into Standby. Click Apply.

Don’t worry if the connections don’t go into standby, just make sure that FOO is applied. Sometimes with multiple mapped disks this can happen.

Now is you run the same mpclaim command your LB Policy should be changed to FOO (Fail Over Only). You will need to do this for each mapped disk.

To change the Load Balancing policy to FOO run mpclaim with the -L and -M switch.

mpclaim.exe -L -M 1

The one at the end is indicative of a FOO LB policy, if a connection fails it will immediately fail over to the next one. This is for always on high demand systems.

Now if you run the -s -d switches you should see FOO under the LB policy.

Now go into Control Panel > Administrative Tools > Computer Management and bring the iSCSI disk online and format them to NTFS. I had an instance where the disk wouldn’t come online even when I brought it online. If this is the case resize your LUN disks, they are too large.

Mpclaim determines the policy for the iSCSI connection. For more information on mpclaim go to Microsoft’s website and user the following reference https://technet.microsoft.com/en-us/library/ee619743(v=ws.10).aspx

https://technet.microsoft.com/en-us/library/dd851699(v=ws.11).aspx

iSCSI and VHD/VHDX volume size restrictions.