Analyzing esxtop results
The data collected by esxtop is processed as rates. When you run the esxtop command without any parameters, it presents the server-wide and individual statistics of a VM, the resource pool, the CPU utilization, and a world. A world is a technical term, like a running process in VMware terms, that represents a single VMkernel schedulable entity or a process or processes running on VMkernel. As we have a total of 369 worlds (see Figure 2.1), it is called group of worlds. These worlds can belong to a group of idle worlds, group of system worlds, group of helper worlds, or another group of worlds.
Understanding CPU statistics
I will explain the statistics found in Figure 2.1. The first line you can find in esxtop or vtop starts with the current system time. The uptime indicates how long the host is up. As you can see in the preceding figures, our host is up since the last 431 days. The next is the number of worlds, which in our case is 369.
The next option indicates the number of virtual machines the host is currently hosting. In the Figure 2.1, it indicates that the host is currently hosting only a single virtual machine. The value 16 vCPUS indicates that the host has 16 virtual CPUs, and the last line indicates the average CPU load. Currently, our load average is 0.02, but if it goes to 1.00 it means that the physical CPUs of our host have been fully consumed. In the same way, if it indicates average CPU load at 0.75, it means 75 percent of the CPU is being utilized by the vSphere host system. If the load average is greater than the CPU cores you have, it indicates that the CPU resources of your vSphere host are overloaded and requires the CPUs to be doubled up.
The PCPU section indicates the percentage of individual physical CPUs. You can see the usage of 16 physical CPUs in Figure 2.1. The value of Avg. 05 PCPU shows the total average percentage in all of the 16 CPUs. The ideal usage of CPUs should be up to 80 percent. CPUs are considered overloaded if the average percentage of PCPU reaches 90 percent, but not for all organizations. However, usage of 90 percent and above is clearly a warning that CPU resources are going to be overloaded. If a vSphere host is installed on a machine enabled with hyper-threading, the PCPU field also displays information about the logical CPU (LCPU) usage. If a vSphere host in a cluster fails, vSphere HA will try to start the virtual machines for the other available vSphere hosts, but if the CPU resources on the vSphere hosts are already exhausted, your virtual machines will not be powered on.
The rest of the metrics are presented in the table in the next section, and most of them are self-explanatory. The CPU panel can be customized using single-key commands as you do in the top tool in Linux or Windows.
Enabling more esxtop fields
You can perform the following steps in order to choose more fields for the CPU screen:
- Press f to enter another screen called the Current Field Order screen.
- You can toggle any of the fields you would like to display. The field order starts from A to J.
- For this exercise, press I to choose CPU Summary Statistics.
- Press Esc to get back to the CPU statistics screen. You will see that the number of columns has increased.
The same procedure can be adopted for other statistics screens, for example memory, network, or disk screens. You can also use e for extended statistics about a group. Press e while esxtop is running in the interactive mode. Enter the Group ID (GID) for which you would like to see the extended statistics, and esxtop will display the extended information about that resource group.
Pressing U will sort all the resources by utilization; pressing R will sort all the resources by %RDY (see Table – 2.1 CPU Metrics) state. Pressing V will display resources relevant to virtual machines, and N will sort them by the GID column.
Table – 2.1 CPU Metrics
