Collecting performance snapshots using vm-support in ESX and ESXi
search cancel

Collecting performance snapshots using vm-support in ESX and ESXi

book

Article ID: 326320

calendar_today

Updated On:

Products

VMware vSphere ESXi

Issue/Introduction

This article provides information on collecting performance snapshots in ESX and ESXi.

Environment

 

VMware ESXi Server 7.x.

VMware ESXi Server 8.x.

Resolution

Performance snapshots are continuous snapshots of proc nodes, collected at defined intervals for a specified period. Technical support may ask you to send performance snapshots to help troubleshoot a performance issue in response to a service request that you opened.

HTTP-based download of vm-support output from an ESXi:

Using any HTTP client:

https://ESXi_host_FQDN/cgi-bin/vm-support.cgi?performance=true&interval=5&duration=300

After the log bundle is collected and downloaded to a client, upload the logs to the SFTP/FTP site. 

Command line based

To collect the performance snapshots, run this command:

  • vm-support -p

You may specify the collection duration and interval in seconds using these command-line options:

  • vm-support -p -d duration_in_seconds -i interval_in_seconds

To collect only performance snapshots from an ESXi host, run this command:

  • vm-support -p -a PerformanceSnapshot:vsi -d duration_in_seconds -i interval_in_seconds

If an interval is not specified, vm-support automatically sets the interval period to two times the amount of time taken to collect a single snapshot. For most cases, VMware recommends that you set the interval period to 10 seconds.

Note:

1. To run the vm-support command, log in to your ESXi host with root privileges using the console or an SSH session.

2. The collected snapshots are bundled with regular data collected by vm-support and a tar archive is created in /var/tmp/esx-20##-01-26--20.46.3236587.tgz. To avoid space issues, the log files can be redirected to datastore by specifying the path.

  • To redirect to a specified location include this parameter in the vm-support command: -w /vmfs/volumes/storage

3. If you want to collect only performance snapshots without the regular data, substitute the -s with an uppercase -S.

4. It is important that you use the vm-support script to capture performance snapshots when your system is experiencing a performance problem. The script cannot collect useful data otherwise. If the snapshot collection interval overlaps with a time when the virtual machine is not experiencing a            performance problem, the result may skew the data collected and can make analysis difficult.

5. To download the vm-support script, see "Collecting diagnostic information for VMware ESXi". This article also contains related information about collecting debug information for ESXi systems.

6. The interval parameter is a pause inside each snapshot, not the time between the start of one snapshot and the next. Each snapshot takes the interval plus the time needed to collect its data, so snapshots are spaced further apart than the interval value.
 
7. The expected number of snapshots is approximately:
 
   snapshots ≈ duration / (interval + collection time)
 
   Collection time depends mainly on host size (number of CPUs, worlds, devices and paths) and also on load. On large hosts it can be 100 seconds or more per snapshot, so far fewer snapshots than duration / interval may be collected. For example, with interval=10 and a collection time of 130              seconds, duration=300 produces 3 snapshots. To collect more snapshots, increase duration.
 
8. The duration is checked only before each snapshot starts. A snapshot that starts just before duration ends still runs to completion. After the snapshots, the full log bundle is collected, so the total request time can be much longer than duration. Set a client timeout well above duration. On large hosts      the full request can take 20 minutes or more.
 

Additional Information

 

Powered by Wolken Software