Hi,I have a few questions about how machines are distributed across hypervisors.
Is there any documentation on how machines are distributed across hypervisors besides http://project-fifo.net/display/PF/Packages ?
In my test environment I have three hypervisors that support kvm & zones and one that supports zones.
Out of the box with one package all machines are created on one hypervisor.
Without any rules defined to influence hypervisor selection I'd expect a least-loaded host or random distribution across possible hypervisors. E.g. In my case a kvm machine could land on one of three hosts while a zone could land on one of four.
Is there a way to set a random distribution?
Is there a recommended way to ensure that machines are provisioned on different hypervisors for redundancy?
The only way I see is to group my hypervisors into redundancy groups using characteristics and then use rules on packages.
If I have a three-server MySQL cluster (one master, two slaves) it seems like I'd need three packages that only differ in the rules for hypervisor selection.
Package Generic: random server assignment
Package Generic-Redundancy-Group-A: rule to restrict to servers in redundancy group A.
Package Generic-Redundancy-Group-B: rule to restrict to servers in redundancy group B.
This seems overcomplicated. Am I overthinking this?
Ideally I'd only need one package for each resource allotment (CPU, RAM, storage) and then influence hypervisor selection when I provision a new machine by selecting the specific hypervisor or putting in a rule like "any machine except the one that UUID-EXAMPLE is running on".