As every esx engineer knowns its important to pay close attention to the core four components in a Vmware Infrastructure 3. For the most important counters see table below,referring to online documentation on performance monitoring and analysis for more information on these counters and how to interpret them.

Subsystem Esxtop Counter VirtualCenter Counter
CPU %RDY

&USED

 Ready

Usage
Memory %ACTV

SWW/s

SWR/s

 Active

Swapin

Swapout
Storage ACTV

DAVG/cmd

KAVG/cmd

 Commands

deviceWriteLatency and deviceReadLatency

kernelWriteLatency and kernelReadLatency
Network MbRX/s

MbTX/s

 packetsRx

packetsTx

Table 1 core fore counters

How to Interpret the disk counters

The following is NOT A GOLDEN RULE, but in general terms (own experience and from several vmware support cases)

In my esx career network is never beeing the problem.

Cpu and memory on a physical esx host can be “easily” monitored and can be “easily” interpret to be a bottleneck.

Because the virtual machines run on a san/nas, its important to monitor the disk counters to see if there is a bottleneck.

Below the values of the disk counters and what is good and what is really bad.

Disk counter Very good Busy Really bad
Davg/cmd 100- 200+/300+ 800+
Kavg/cmd It doesn’t change much. Normal values are 0.0X

esxtop

The drawback of esxtop is that its only realtime and that you cannot look back in time and that it is per esx server. So I thought that it would be very nice to use virtualcenter for the monitoring of these counters so you can look back in time and can monitor all the esx servers through a central management console.

virtualcenter

I found out that this is possible to monitor specific counters with virtualcenter but that this has also some drawbacks.

  • You have to change virtualcenter server setting which will let the vc database grow (depending on your environment)
  • you have to make custom_charts in the vi client per esx server because the default charts do not monitor the most important counters
  • it looks like the custom_charts are getting saved on the local vi_client or user profile (I did not found out where or how)

although that there are drawbacks, you have to set this up once, below the steps how.

how to setup the performance charts

STEP1 first of all you have to change the virtualcenter management server configuration and change the statistics level of each statistics interval otherwise you can only monitor your custom_chars real time, and that’s not what you want :) .

In the vi client go to administration-“virtualcenter management server configuration”-statistics

default is statistics level 1, change it to 3 or 4 !!!! BE AWARE that changing these values on a large esx environment can have a big impact on the virtualcenter performance and the virtualcenter database. !!!!

STEP2 Making the custom charts in virtual center

click on a esx host go to performance tab and click change chart options

click on one of the core four components you want to make a custom chart of , change the metrics to the right metrics as mentioned in table1 and click save changes and give the custom chart a name

As you can see below you now have custom charts, which you can use to look back in time. As mentioned the big drawback is that you have to do this for every esx host and saved somehow on the client were the vi client is installed :(