I recently had a VNXe 3150 installed and configured. Along with that purchase we had gone ahead and gotten a new Dell PowerConnect 5524 to go with our Dell PowerConnect 5424 so that we could have more redundancy. Recently, I got around to configuring the multipathing in our vSphere 5.1 environment. I thought that I would be able to unplug an ethernet cable from each SP and plug it into the new switch (after it was configured) and then unplug a link on our host and plug it into the new switch. However, this disconnected the datastore, which was not desirable in our environment.
After doing a ton of digging and thinking, I figured the issue was with the configuration of our SAN. While looking into the multipathing on our VMware side, I noticed that the iSCSI adapters were mapped to the same IP addresses.
As you can see, both were pointed to 192.168.130.8. So, when I disconnected the cables indicated below, depending on the cable, it would have disconnected the datastore.
In order to resolve this, I had to work with both Unisphere software and VMware vCenter. First, I noticed in Unisphere that I had 4 iSCSI Servers all pointing to a different IP Address.
So, I needed to resolve this. I added in two more iSCSI servers, SPA and SPB. In SPA I added in two IP Addresses that would point to the eth interfaces. Below is the SPA setting.
Once I got SPA made, I went ahead and created SPB.
I then had to move over to vSphere and configured the iSCSI port binding. I followed this excellent video:
After getting the port binding setup, I then had to add the new iSCSI targets using the ‘Dynamic Discovery’ on the edit iSCSI Initiator properties. Once those got added, I then had to configure my EMC SAN datastores to point to the new IP Addresses. This required me to create new datastores on the SAN. Since I didn’t have very much extra space and am not licensed to use Storage vMotion, I had to take quite a bit of downtime. I ended up changing one datastore at a time by migrating all VMs off of that datastore, deleting it in Unisphere and then recreating the datastore, making sure to select SPA or SPA iSCSI Server. Once I did that, I presented the LUN to my hosts, which picked up the datastore right away.
Each datastore was upgraded to VMFS 5 and the paths were setup to ‘Round Robin’. Making sure that each path was ‘Active’.
I then continued moving around the VMs until I got to vCenter, which was kind of tricky… http://goo.gl/uuHCJn
Once all of the datastores were re-added and the settings were configured properly, I disconnected the networking cable from my host (that wasn’t running VMs) and connected it to the new switch. It all worked and everything came back to Active/Active!
All three hosts are configured exactly the same are all working as required.