IBM Support

Forcing crash on hung Linux VMs

Question & Answer


Question

How can I force a crash against a Linux VM that appears to be hung or frozen and no longer responding to external input?

Cause

Once a VM stops responding to input, it can be difficult to diagnose what exactly the Linux Guest OS is doing. One method to accomplish such investigation is to force a "crash" of the Guest OS and obtain a recording of the Guest OS state at the time of the "hang".

Answer

Once the Linux Guest OS is configured for a crash facility, such as kdump, then an NMI can be sent to the Guest OS via the Service Console command line of the ESXi host.

The method for generating the NMI to the virtual hardware is documented at under the "vmdumper" section:
Crashing a virtual machine on ESX/ESXi to collect diagnostic information
http://kb.vmware.com/kb/2005715

This method has been shown to work effectively with RHEL5 Guests.

Note
====
Problems have been found with accomplishing this using SLES10 Guest OS and vmxnet network drivers. Those problems do not yet have a solution or work around at this point. The kdump may or may not complete successfully in such a case, as the vmxnet drivers are not integrated into the SLES kdump kernel.

[{"Product":{"code":"SSCLB3","label":"VMware Solutions"},"Business Unit":{"code":"BU053","label":"Cloud \u0026 Data Platform"},"Component":"ESX","Platform":[{"code":"","label":"VMWare"}],"Version":"4.0;4.1;5.0;5.1;5.5","Edition":"","Line of Business":{"code":"LOB66","label":"Technology Lifecycle Services"}}]

Document Information

Modified date:
28 January 2020

UID

isg3T1011265