fixed graphing failures for VM deployed on legacy Centos 6.9 virtualization
platforms (script uses the Proxmox entries), in the case of VM domain names
beggining with digits (first field name character must be [a-ZA-Z_])
used "Validate field names" Python stanza from documentation ("How to write
plugins", "Notes on field names") to improve the clean_vm_name function, thus
munin kvm graph creation succeeded for VM with names starting with digits,
like "150-121-Apache", deployed on Centos 6.9 virtualization platforms.
fixed graph failure for VM names starting with digits
used "Validate filednames" Python stanza from docs to improve
clean_vm_name functions, helping graph creations for VM with names
starting with digits, like "150-121-Apache".
remember: each data source in a plugin must be identified by a field
name (in this case VM names). The characters must be [a-zA-Z0-9_],
while the first character must be [a-zA-Z_].
The network interface parser of this plugin was overly specific before.
It relied on a specific format of the arguments handed over to kvm while
starting the VM. For example the following format was usable:
... -netdev tap,ifname=foo,...
But kvm/qemu support a variety of ways for configuring network
interfaces via the commandline. E.g. libvirt does not use the "ifname"
parameter above. Thus VMs running on a host controlled via libvirt
cannot be tracked with this plugin.
This limititation is now clearly documented in the header of the plugin.
The old parser seemed to rely on a simple "-name foo" argument format of
kvm/qemu. The changed parser also accepts the following formats:
* name,foo=bar,baz=bot
* guest=name,foo=bar
