Showing posts from September, 2020

HPE Power Profile Modes explained and best practice for HPE Server

 Today I discoverd a VM that was running slower on one host than on the other. This was due to Power Profiles misconfigured. Having them set to "Balanced Power and Performance" lead to only C-States are exposed to the Hypervisior and thus leading to lower control over CPU clock speeds. The bad thing about this is that there is no default option for this but you have to select the Custom Profile and have Power Regulator set to "OS Control Mode". This will expose P- States to the hypervisor. The cool thing about it, it´s also available via iLO at least on Gen9 and Gen10 but you have to reboot anyway for changes to take effect. Having now both ACPI states available to the hypervisor gives you now the option to either select High Performance that will limit the ability to leverage turbo boost but having consistent performance or even better selecting "Balanced Performance" that will unlock turbo bin´s but more about this is perfectly written down here starting

"Really" enabling promiscous mode for vSphere VMs (VMWare KB 59235)

 We had this situation twice now where just enabling promiscous mode on a dvSwitch port group was not sufficient to the application. There is one setting that need to be enabled as well especially when using more then one uplink (what you should have for redundancy) except when using LACP. This is documented in following KB Duplicate Multicast or Broadcast Packets are Received by a Virtual Machine When the Interface is Operating in Promiscuous Mode (59235) You may set this via PowerCLI to save the hassle for ssh´ing to all needed hosts: $options = @{option="/Net/ReversePathFwdCheckPromisc";intvalue=1} $esxcli = get-vmhost "YourHostnameHere" | get-esxcli -v2 $esxcli.system.settings.advanced.set.Invoke($options)