Description Philipp Hahn 2013-12-12 15:18:05 CET

/usr/lib/univention-virtual-machine-manager-daemon/uvmmd-check.sh checks for "uvmm groups" to return any data, which communicates with *UVMMd*, which is running on the management node.

If no data is returned, it restart "libvirt-bin" AKA *libvirtd*, which is supposed to be running only the virtualization hosts.

Only when the management node is also a virtualization host does the script do anything useful.

Fix:
-                invoke-rc.d libvirt-bin restart >/dev/null 2>&1
+                invoke-rc.d univention-virtual-machine-manager-daemon restart >/dev/null 2>&1

In addition to that a similar script might be used on the virtualization hosts to check if libvirtd is still working:

#!/bin/bash
# Check for functional libvirtd
exec >>"/var/log/univention/virtual-machine-manager-daemon-errors.log" 2>&1
if [ -f /proc/xen/privcmd ]
then
  uri="xen+unix:///"
elif [ -c /dev/kvm ]
then
  uri="qemu:///system"
else
  exit 0
fi
if sv status /etc/runit/univention-libvirt | grep -q ^run
then
  tmpfile="$(mktemp)"
  trap "rm -f '$tmpfile'" EXIT
  virsh -c "$uri" list >"$tmpfile" &
  pid=$!
  sleep ${uvmm_check_timeout:-5}s
  if [ ! -s "$tmpfile" ]
  then
    kill $pid
    wait $pid
    invoke-rc.d libvirt-bin restart
  fi
fi
exit 0