Having a top-down monitoring tool to survey your virtual landscape is valuable, but when something breaks, a VMware administrator should also have a ground-up understanding of how different technology pieces interact to uproot performance issues.
Knowing how the assignment of resources to a VM affects the host, networking and storage components can help prevent issues when demand begins outstrip an application's ability to deliver in a timely fashion.
To help VMware administrators better understand the basics of the inner workings of vSphere, Jonathan Frappier, an IT professional with 15 years of experience and blogger at www.virtxpert.com, wrote VMware vSphere Resource Management Essentials.
SearchVMware.com spoke with Frappier about tools and techniques admins can use to diagnose problems within their vSphere environment.
What are some common mistakes you've seen when a VMware administrator tries to fix a performance problem?
Jonathan Frappier: I've seen many people just assume CPU and memory is the cause of all performance problems, and the first step someone took was to add more RAM or vCPU. However, by examining the application and OS you can see there is plenty of CPU -- from a guest perspective -- and RAM. Then it turns out the problem was with a software configuration rather than a VM.
We were able to be lazy in terms of application troubleshooting for a few years. Hardware was well beyond what a single server could consume, so a Web server with dual quad-core processors and 16 GB of RAM was great, but that made some of us -- me included -- a bit lazy.
What are some underused ways to give a VMware infrastructure a boost?
Frappier: My first step would be an assessment of the infrastructure using something like vCenter Operations Manager, Veeam One or Dell Foglight. There is probably a good chance there are overprovisioned and abandoned VMs. This is a good broad approach. When you finish running a diagnostic test, you can narrow in on specific aspects.
Not everyone likes VMware's decision to phase out the vSphere desktop client. What are your thoughts about the shift to the Web Client?
Frappier: We tend to not like change. The desktop client is like our favorite blanket. However, having used the Web Client almost exclusively, I've started to forget how to do things in the desktop client.
Of course, if you really don't like the Web Client, you can always live in PowerCLI!
Are there any Flings or third-party utilities you recommend?
Frappier: It depends on your environment and background.
If you are a Windows shop, then PowerCLI and WebCommander are great starts.
WebCommander is a Fling that allows you to publish PowerCLI/PowerShell scripts to end users to perform a task. If you've built a solid set of scripts for troubleshooting and maintenance, you could expose them in a self-service type portal for people to run without IT intervention.
If you are a Linux shop, I'd lean toward Ansible. I've seen configuration problems fixed across multiple servers in just a few minutes with Ansible vs. having to log into each server. Automation is key.
What additional features or improvements would you like to see in the next version of vSphere?
Frappier: A few that come to mind are squashing bugs that have existed in the Web Client for some time, especially if this becomes the de facto client.
Deploying an OVF is the one Web Client bug I run into most often. It's fixed in vSphere 5.1, but still potentially broken in vSphere 5.5. The Knowledge Base article doesn't say why the error occurs. The only workaround is the desktop client. Those types of workarounds will need to be eliminated.
Speaking of VSAN, I've heard some complaints about the pricing. Is it something viable for SMBs?
I think given the ease of use of VSAN -- both in implementation and expansion -- that would offset any cost difference between it and an equally powerful SAN/NAS. Of course, I'd suggest to anyone that if you are paying list price for something, your sales rep doesn't like you very much, so ask for a discount.