Monitoring VAAI with ESXTOP

User Guides for VMware Solutions

Audience
Public
Content Type
User Guides
Source Type
Documentation

The simplest way to specifically monitor VAAI activity is through the ESXi performance monitoring tool ESXTOP. ESXTOP is both a real-time monitoring tool and a time-period batch performance gathering tool.

To monitor activity in real-time, log into the ESXi shell via SSH (SSH is not enabled by default—this can be done through the console or from within the vSphere Web Client in the security profile area) and run the ESXTOP command.

Figure 20. Starting ESXTOP

ESXTOP offers a variety of different screens for the different performance areas it monitors. Each screen can be switched to by pressing a respective key. See the table below for the options.

Key

Description

c

CPU resources

p

CPU power

m

Memory resources

d

Disk adapters

u

Disk devices

v

Virtual machine disks

n

IP Network resources

i

Interrupts

Figure 21. ESXTOP columns

In most cases for VAAI information the disk devices screen is the pertinent one, so press “u” to navigate to that screen. By default, VAAI information is not listed and must be added. To add VAAI information press “f” and then “o” and “p”. To deselect other columns simply type their corresponding letter. Press enter to return.

The CLONE or MBC counters refer to XCOPY, ATS refers to Hardware Assisted Locking, ZERO refers to WRITE SAME and DELETE refers to UNMAP. If a collection of this data for later review is preferred over real time analysis esxtop can be gathered for a specific amount of time and saved as a CSV file.

Batch mode, as it is referred to, takes a configuration of ESXTOP and runs for a given interval. This can create a tremendous amount of data so it is advised to remove any and all counters that are not desired using the “f” option and saving the configuration with “W”.

Once the configuration is complete (usually just VAAI info and possibly some other storage counters) ESXTOP can be run again in batch mode by executing:


esxtop -b -d 5 -n 50 > /vmfs/volumes/datastore/esxtopresults.csv

Make sure the csv file is output onto a VMFS volume and not onto the local file system. Putting it locally will often truncate the csv file and performance data will be lost.

The “-b” indicates batch mode, “-d” is a sample period and “-n” is how many intervals should be taken, so in this example it would take counters every 5 seconds, 50 times. So a total capture period of 250 seconds.