Right-sizing VMs isn't an easy task; the process requires coordination with application owners to shut down VMs, which can cause workload delays and system outages. The VMware VM Desired State Configuration Fling automates the desired state configurations of VMs with an API to properly right-size VMs, which can mitigate application performance issues and resource contention.
In his VMworld 2021 session called "Automated Rightsizing of VMs with VM Desired State Configuration," speaker Steve Tilkens, customer experience architect at VMware, shared an in-depth demo of VMware's VM Desired State Configuration (VMDSC) Fling and its ability to improve and simplify VM right-sizing processes.
VMDSC Fling seeks to eliminate common VM right-sizing challenges
Incorrectly sized VMs can lead to a host of issues, such as application performance issues, system outages and unnecessary resource consumption. You must first shut down a VM to right-size it. This can lead to limited remote access, long startup times and workload disruption.
"Right-sizing VMs is a crucial step in running efficient infrastructure. From my experience in all verticals and customer types, two factors exist when running right-sizing initiatives in real-life: organizational -- my application requires this amount of memory and we never touch a running system -- and technical burdens, [such as] 'how do I size [my application] properly and when can we execute that change technically?'" said Fabian Lenz, partner and virtualization and cloud expert at Comdivision consulting GmbH.
VMDSC Fling's goal is to automate and implement changes made to the VM CPU or memory desired state configuration through an API upon the VM's reboot. This eliminates the need to schedule VM downtime with application owners and prevents unplanned system outages.
"What we wanted to do was create a mechanism within ESXi or vCenter for VM administrators to configure the desired state of virtual machines. And the key point is that configuration will not take effect until the VM is rebooted," Tilkens said in the session.
If you modify a VM's CPU or memory configuration to right-size the VM, the VMDSC Fling ensures those modifications aren't implemented until a planned VM reboot occurs, such as a security patch or Windows update.
VMDSC Fling requirements
Before you install and deploy VMDSC Fling, you must have at least vCenter 7.x or above as well as a vSphere Service Account with the following enabled permissions:
VMDSC Fling uses an API to deliver VM right-sizing configurations within vRealize Operations to vCenter.
"We wanted to provide [VMDSC] through an API, that way you could tie it in with other automation systems and management systems like the vRealize Suite, for instance. Things get a lot more powerful when you can automate," Tilkens said in the session.
VMware delivers VMDSC as an OVA appliance, which is a VM that you can import into your VMware infrastructure. Once you install the VMDSC OVA, you can then direct VMDSC to vCenter and assign any appropriate credentials.
Learn how to setup the VMDSC Fling
To understand how VMDSC Fling can right-size VMs, imagine you have a few workloads within vRealize Orchestrator. You can integrate those workloads into the vRealize Orchestrator Management Pack via vRealize operations, as demonstrated by Tilkens in the session.
Within the vRealize Operations interface, there is a Rightsizing tab that provides an overview of oversized VMs within your infrastructure available for right-sizing. You might have a cluster of workloads and within that cluster resides a SQL server with 12 GB of CPU and 16 GB of memory.
VRealize Operations could flag that server as oversized and recommend removing 6 GB worth of CPU and 8 GB worth of memory from the desired state configuration to right-size the SQL server based on its current demands.
If you decide to right-size that specific workload, you can select the SQL server and navigate to the Actions tab, which then offers a host of new options provided through the vRealize Orchestrator Management Pack. There you will find the option VMDSC - Set VM Desired State - Automatic. Once you select this option, vRealize Orchestrator reaches out to vRealize Operations to obtain the desired state configuration for the SQL server. VRealize Operations then uses the VMDSC API to schedule a planned downtime to implement the set desired configuration.
You can take it one step further and automate right-sizing recommendations within your cluster. In vCenter, if you click on the desired cluster, you can select the VMDSC - Rightsize All VMs within Cluster option under the Actions tab. VRealize Orchestrator then goes through every VM within the cluster, communicates with vRealize Operations to find any configuration recommendations and automatically schedules downtimes for those changes.
"One of my customers does a quarterly right-sizing process where thousands of VMs are shut down [and] right-sized based on vRealize Operations data. [They then] power the virtual machines back on and verify their functionality. This procedure has been automated in [such a] way that complexity wouldn't be required anymore with the Virtual Machine Desired State Configuration that schedules the virtual hardware change after the next power-cycle of the VM. A huge benefit [in regards] to self-services and simplified IT," Lenz said.