Skip to main content
Hello,



I am an engineer with a security software vendor. We have a mutual customer that seems be experiencing system hangs on their vms in use on their Nutanix Acropolis platform. Admittedly I am not very familar with Nutanix as I mostly work with Vmware. These hangs are reportedly occurring after they upgraded our security product.



Vmware has a process where we can create snapshot files and convert them into a memory dump file. This is helpful for our troubleshooting effort when normally you cannot take any action within the vm image to force a dump.



I am trying to find out if the Nutanix Acropolis infrastraucture has any comparable capability. I will be suggesting that our mutual customer reach out to you as well, but am trying to inverstigate capability in parallel.



Your assistance and advisement is welcome.



David Alexander
I'll ask internally for you as I don't know but what OS is your appliacne running?
Also check out KB 3108







check these logs and contact Nutanix Support or your Alliance memeber.


  • ~/data/logs/acropolis.out on the Acropolis Master
  • QEmu log on the host which runs the Guest VM (UVM logs in the AHV host --> /var/log/libvirt/qemu/{vm_uuid}.log)
  • ACLI Task log

You can try to configure the guest to create a dump file on a persistent disk. After that, when the guest hangs, inject an NMI to it using virsh inject-nmi (https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Virtualization_Deployment_and_Administration_Guide/sect-Generic_Commands-Injecting_NMI.html). If the guest OS is Windows 2012 or later, it’s configured for NMI already. If it’s Linux, I’m not sure, but I suspect it must be setup.

Hey there, I know this is three years old, but I was wondering if there were any updates on this.  Specifically, VMWare has a process that allows you to create a vmcore from a .vmss file for Linux systems (or Windows systems I presume, although I do not work with Windows) without having to manually configure kdump and panic-on-nmi on the system itself, and then have AHV send an NMI.

 

Is this something AHV can do?  I don’t see KB 3108 as existing anymore.  This is extremely useful for catching guest VM’s during problems when potentially even kdump can’t work, as you just dump the memory of the guest to a file which can be analyzed by a kernel engineer using kgdb or the crash utility.  KVM can do this as well with `virsh dump`.

 

Thanks!