When it comes to designing storage for virtualization, there is one important step that happens in every environment, it seems. It is the whiteboard step where each ESX(i) server is drawn up to have two or more HBAs to two or more different switches to a storage controller with two or more interfaces. This also occurs with iSCSI with multiple initiator interfaces going to different switches to the storage controller. This is shown in Figure A below, and can be tweaked and changed in many ways with different storage products or network configurations:
The design up front is very important, but it is also critical to check inside of the ESX(i) host to see that the expected number of paths are present on the hosts. But even more importantly, you want to see if any are listed as broken. Within the vSphere Client, we can easily see the status of the paths. Click on the host and then the configuration tab, select a datastore, and the path information is shown. Figure B is an example of a host with two paths to a datastore:
Click to enlarge.
But the catch here is that each host has its own path configuration. And while the datastore properties on one host may be correct, there is no guarantee that the next host in the cluster is configured as you expect it to be. The properties of the datastore can show which path is active and change the multipathing policy, but this is a configuration that is live for each host in a cluster.
The vSphere Client makes it easy to take a quick look at a host’s configuration for paths to a datastore. What tricks have you used to check to ensure that each host has the requisite and planned number of paths to datastores? Share your comments below.