On Tue, Aug 27, 2013 at 09:09:25AM +0200, David Weber wrote:
Hi,
we try to use vcpu pinning on a 2 socket server with Intel Xeon E5620 cpus, HT enabled and 2*6*16GiB Ram but experience problems if we try to start a guest on the second socket: error: Failed to start domain test error: internal error: process exited while connecting to monitor: kvm_init_vcpu failed: Cannot allocate memory
Libvirt version 1.1.1 Linux 3.11-rc7
Because I coudn't find any other service which allowed a 7M file upload, I put the log file and everything else which could perhabs be relevant into a github repository: https://github.com/David-Weber/vcpu-pinning
When we try to start a guest on the first node it runs fine: <vcpu placement='static' cpuset='0-3,8-11'>4</vcpu> <numatune> <memory mode='strict' nodeset='0'/> </numatune>
Starting it on the second node fails <vcpu placement='static' cpuset='4-7,12-15'>4</vcpu> <numatune> <memory mode='strict' nodeset='1'/> </numatune>
Even more strange, starting it with the CPUs of the second node and the memory of the first node works: <vcpu placement='static' cpuset='4-7,12-15'>4</vcpu> <numatune> <memory mode='strict' nodeset='0'/> </numatune>
The log file contains these three cases.
Using the placement='auto' parameter leads to the same problem. If numad return the second node, the guest won't start.
Is this a configuration, a libvirt or a cgroup problem? :)
With mode='strict' you are telling QEMU that if it can't allocate memory from the requested node, it should fail. Is it possible that some of your Numa nodes have insufficient memory free ? The combination of 'virsh capabilities' output and the results of 'virsh freecell NODENUM' for each NUMA node will give an indication of the allocation state. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|