Updating instance_uuid while scheduling a bulk of VMs
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Opinion
|
Wishlist
|
Unassigned |
Bug Description
This problem concerns custom scheduler filters and weighers which can be added by user in order to specify his own scheduling algorithm.
During scheduling of a single VM, it's UUID can be retrieved from filter_properties:
"request_spec": {
"block_
}
]
}
However, when a bulk of VMs is scheduled, custom filters or weighers are not able to read UUID of currently scheduled VM. "instance_uuid" in filter_properties is always the same. Problem is still present on Icehouse and nova's master.
Passing instance_uuid of a VM that is currently filtered or weighed could be done inside _schedule() method:
class FilterScheduler
...
def _schedule(self, context, request_spec, filter_properties, instance_
...
for num in xrange(
# Set the "instance_uuid" in filter_properties.
# It is accessible from instance_
Changed in nova: | |
assignee: | nobody → Mateusz Blaszkowski (mateusz-blaszkowski-8) |
status: | New → In Progress |
Changed in nova: | |
status: | In Progress → Confirmed |
Changed in nova: | |
assignee: | Mateusz Blaszkowski (mateusz-blaszkowski-8) → nobody |
Changed in nova: | |
status: | Confirmed → Opinion |
Fix proposed to branch: master /review. openstack. org/103902
Review: https:/