Cluster nodes crash shortly after starting install of VMs

[dgeist]

Hi, I'm running a 3-node PM3.4 installation on top of an Ubuntu 14.04 workstation/KVM. This is for demonstration purposes, the successful result of which would lead to deployment on bare metal.

The problem is that every time I start to install a VM, the node on which it's started locks up, reports CPU faults on console, the crashes entirely. The cluster is running a basic Ceph setup for block storage (32 Gb total storage). Each VM node has 8 processors and 4 Gigs of RAM, and a "Virtio" system disk along with an emulated SCSI disk for the CEPH OSD.

The client VMs I'm trying to install are Ubuntu 14.04 and Centos 6.6.

Any thoughts as to what could be causing the instability?

Dan

 

[wolfgang]

Hi,

You have enable nested visualization?
Can you send the CPU faults?

 

[dgeist]

Hi,

You have enable nested visualization?
Can you send the CPU faults?

Hi. I have enabled the nested virt feature on my libvirt hypervisor (and have nested other virtualized hypervisors on it previously). Do I need to enable nested virt in the 3 virtual Proxmox hypervisors as well? I saw a note somewhere that using virtio in a nested client can cause problems. Is this the case? Also, I can't get any meaningful logging out of proxmox, even on the console. Other than usual linux syslog, where does one even look for the problem. It's a bit of a black box. Things just stop responding and the other two nodes in the cluster start reporting ceph OSD errors, etc.

Dan