Moab Viewpoint
Node Policy Types

Node Policy Types

Allocation Policy

Policies
Click to enlarge

The Allocation Policy editor allows you to edit the allocation algorithm associated with Moab scheduling. The Allocation Policy GUI includes the ability to select from various algorithms. If the Priority algorithm is selected, the Priority Function editor appears. Using node allocation priority, you can specify several priority components for compute nodes. The Allocation Policy editor permits simple editing of these coefficients. The node allocation priority function specified in the GUI is done on a cluster-wide basis. Node-by-node allocation is permitted through the moab.cfg file.

Restart for Changes to Take Effect

For Allocation Policy changes to take effect, a restart of Moab is required. When attempting to save, a dialog displays and gives you the option to restart Moab or save the changes without a restart. If Save is selected, the changes do not take effect until a restart occurs.

'Save and restart' dialog

VM Policies

The VM Policies page lets you set policy options for VM migration, green migration, Reservation Priority, and Overcommit Migration.

VM Migration: Enables migration of VMs within the entire system.

Green Migration: Places as many VMs on as few machines as possible.

Reservation Priority: Asserts that when one or more VMs resides on a machine, and a reservation shows up on that machine, the VMs should be moved off of that machine for the duration of the reservation.

Overcommit Migration: Allows VM resource consumption to go beyond the resources on the actual machine. The following resources can have an Overcommit Factor to specify how far to allow overconsumption. For example, if Processors is set to 1.5 on a machine that has 8 processors, then VMs will be allowed to require up to a total of 12 processors (8*1.5=12) before the machine is considered "full". Overcommit Threshold is a multiplier between 0 and 1 that tells the system the percentage of the available resources on a node to consume before reporting the node as being full.

Throttling: Stops thrashing during VM migration. Basically, if a VM is migrated to a location that it just left (within less than the specified time), it will not be migrated away from that machine again.

Max Allowed Duration Before Failure: Set the maximum amount of time (in hours, minutes, and seconds) a VM Creation, Migration, or Deletion is allowed to continue before it fails.

Migration Scheduling Intervals: Set an interval for how often a Green Migration and/or Overcommit Migration should take place.