VMKernel Memory Dump is truncated on vSAN ESA configured systems

Products

• VMware vSphere ESX 8.x

Issue/Introduction

When a VMKernel memory dump is generated due to a PSOD (Purple Screen of Death) or a live dump collection, the dump is truncated on a system configured with vSAN ESA.

Environment

VMware vSphere ESX 8.0

Cause

On systems configured with vSAN ESA, the default estimated size for the VMKernel memory dump is underestimated. Consequently, the core dump partition or dump file size may be insufficient to store and dump the memory data.

Resolution

This issue is resolved in VMware vSphere ESX 9.0 and later.

In VMware vSphere ESX 8.0, the issue can be addressed by manually increasing the dump file size from the default setting.

The additional size is calculated as follows:

Additional Size = 6 GB + 0.1 GB * (Total vSAN drive capacity in TB)

(Example) If the default size is 10 GB, and the host uses 20 local SSD drives of 7 TB each:

Additional Size = 6 GB + 0.1 GB * 20 * 7 = 20 GB

New Dump File Size (GB) = (Default Size) 10 GB + 20 GB = 30 GB

(New Dump File Size (MB) = 30 GB * 1024 = 30720 MB)

 

The core dump size can be modified using the following esxcli command:

Create a dump file with the specified size (in MB):

esxcli system coredump file add -d (datastore name) -f (file name) -s (new size in MB)

Verify that the file has been created:

esxcli system coredump file list

Activate the dump file:

esxcli system coredump file set -p /vmfs/volumes/(datastore UUID)/vmkdump/(file name).dumpfile

Note: For more details on the commands, please refer to KB 314320 "Configuring ESXi coredump to file instead of partition."