If you’re DataStores are using Fibre channel storage and you have multple fchba’s connected through to the SAN via a fc switch or two, then it is prudent to optimise the IO potential of all the redundant hardware by changing the fc path policy to “round robin”.
Using vsphere client, connect to the ESXi host / vcentre server
Inventory, Hosts and Clusters, select ESXi host
Configuration tab, Storage, highlight the datastore, click Properties
Click Manage Paths button on the DataStore properties dialog
Change Path Selection to Round Robin (VMWARE) – the default is Most Recently Used (VMWARE)
Wait for the screen to reload, Click Close
Repeat for each DataStore, and then Repeat for each ESXi host.
From experience, it’s not uncommon to not be able to connect to the KVM of a cisco ucs blade. Instead of seeing a Remote console screen, you’ll receive a “Connect failed” or “Request Shared Session” message, with no means of getting to the console.
Within the Service Profile, click on the Server Details tab. From there, click on Recover Server. Select “Reset CIMC (Server Controller)”. Choose Reset KVM Controller. This will kill existing KVM sessions and allow you to start a new session. Resetting the CIMC does not affect data traffic to/from the server NICs (ethernet and HBAs).
Another thing to check in Servers tab, General tab is the Management IP Address setting. If it’s configured to take an address from a pool, check the pool in the Admin tab, Management IP Pool, IP Addresses tab to see what IP’s exist in the range, and whats been assigned.
If a reset hasn’t worked, In the Servers tab, General tab, Management IP Address section, change the IP address from Pooled to Static. Use an IP address from the other end of the range in the pool. Click Save Changes, and try connecting to the KVM again.