openstack/placement
Code Issues Proposed changes

Merge "Add microversion 1.39 to support any-trait queries"

Browse Source
This commit is contained in:
Zuul 2022-02-22 20:23:52 +00:00 committed by Gerrit Code Review
commit ff0f2dda77
11 changed files with 641 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
api-ref/source/parameters.yaml
View File

@ -246,6 +246,15 @@ required_traits_granular:
It is an error to specify a ``requiredN`` parameter without a corresponding
``resourcesN`` parameter with the same suffix.
**Starting from microversion 1.39** the granular ``requiredN`` query
parameter gained support for the ``in:`` syntax as well as the repetition
of the parameter. So::
requiredN=in:T3,T4&requiredN=T1,!T2
is supported and it means T1 and not T2 and (T3 or T4).
min_version: 1.25
required_traits_unnumbered:
type: string
@ -264,6 +273,30 @@ required_traits_unnumbered:
associated via aggregate. **Starting from microversion 1.22** traits which
are forbidden from any resource provider may be expressed by prefixing a
trait with a ``!``.
**Starting from microversion 1.39** the ``required`` query parameter can be
repeated. The trait lists from the repeated parameters are ANDed together.
So::
required=T1,!T2&required=T3
means T1 and not T2 and T3.
Also **starting from microversion 1.39** the ``required`` parameter
supports the syntax::
required=in:T1,T2,T3
which means T1 or T2 or T3.
Mixing forbidden traits into an ``in:`` prefixed value is not supported and
rejected. But mixing a normal trait list and an ``in:`` prefixed trait list
in two query params within the same request is supported. So::
required=in:T3,T4&required=T1,!T2
is supported and it means T1 and not T2 and (T3 or T4).
resource_provider_member_of:
type: string
in: query
@ -325,11 +358,35 @@ resource_provider_required_query:
type: string
in: query
required: false
description: >
description: |
A comma-delimited list of string trait names. Results will be filtered to
include only resource providers having all the specified traits. **Starting
from microversion 1.22** traits which are forbidden from any resource
provider may be expressed by prefixing a trait with a ``!``.
**Starting from microversion 1.39** the ``required`` query parameter can be
repeated. The trait lists from the repeated parameters are ANDed together.
So::
required=T1,!T2&required=T3
means T1 and not T2 and T3.
Also **starting from microversion 1.39** the ``required`` parameter
supports the syntax::
required=in:T1,T2,T3
which means T1 or T2 or T3.
Mixing forbidden traits into an ``in:`` prefixed value is not supported and
rejected. But mixing normal trait list and ``in:`` trait list in two query
params within the same request is supported. So::
required=in:T3,T4&required=T1,!T2
is supported and it means T1 and not T2 and (T3 or T4).
min_version: 1.18
resource_provider_tree_query:
type: string

9
doc/source/user/provider-tree.rst
View File

@ -205,6 +205,11 @@ spec for details. The ``required`` query parameter also supports negative
expression, via the ``!`` prefix, for forbidden traits. If a forbidden trait
is specified, none of the resource providers that appear in the allocation
candidate may have that trait. See the `Forbidden Traits`_ spec for details.
The ``required`` parameter also supports the syntax ``in:T1,T2,...`` which
means we are looking for resource providers that have either T1 or T2 traits on
them. The two trait query syntax can be combined by repeating the ``required``
query parameter. So querying providers having (T1 or T2) and T3 and not T4 can
be expressed with ``required=in:T1,T2&required=T3,!T4``.
For example, let's say we have the following environment::
@ -493,6 +498,10 @@ each optionally prefixed with ``!`` to indicate that it is forbidden.
.. note:: When sharing providers are involved in the request, ``root_required``
applies only to the root of the non-sharing provider tree.
.. note:: While the ``required`` param supports the any-traits query with the
``in:`` prefix syntax since microversion 1.39 the``root_required``
parameter does not support it yet.
Filtering by Same Subtree
=========================

4
placement/microversion.py
View File

@ -99,6 +99,10 @@ VERSIONS = [
# type irrespective of whether the ``consumer_type`` was specified
# in the request. The corresponding changes to ``/reshaper`` are
# included.
'1.39', # Adds support for the ``in:`` syntax in the ``required`` query
# parameter in the ``GET /resource_providers`` API as well as to
# the ``required`` and ``requiredN`` query params of the
# ``GET /allocation_candidates`` API.
]

15
placement/rest_api_version_history.rst
View File

@ -698,3 +698,18 @@ considered to have an ``unknown`` ``consumer_type``. If an ``unknown``
``unknown``.
The corresponding changes to ``POST /reshaper`` are included.
1.39 - Support for the any-traits syntax in the ``required`` parameter
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. versionadded:: Yoga
Adds support for the ``in:`` syntax in the ``required`` query parameter in the
``GET /resource_providers`` API as well as to the ``required`` and
``requiredN`` query params of the ``GET /allocation_candidates`` API. Also adds
support for repeating the ``required`` and ``requiredN`` parameters in the
respective APIs. So::
required=in:T3,T4&required=T1,!T2
is supported and it means T1 and not T2 and (T3 or T4).

336
placement/tests/functional/fixtures/gabbits.py
View File

@ -772,3 +772,339 @@ class LegacyRBACPolicyFixture(APIFixture):
"""An APIFixture that enforce deprecated policies."""
_secure_rbac = False
class NeutronQoSMultiSegmentFixture(APIFixture):
"""A Gabbi API fixture that creates compute trees simulating Neutron
configured with QoS min bw and min packet rate features in multisegment
networks.
"""
# Have 4 trees. 3 trees with the structure of:
#
# compute
# \ VCPU:8, MEMORY_MB:2095, DISK_GB:500
# |\
# | - Open vSwitch agent
# | | NET_PACKET_RATE_KILOPACKET_PER_SEC: 1000
# | \ CUSTOM_VNIC_TYPE_NORMAL
# | \
# | - br-ex
# | NET_BW_EGR_KILOBIT_PER_SEC: 5000
# | NET_BW_IGR_KILOBIT_PER_SEC: 5000
# | CUSTOM_VNIC_TYPE_NORMAL
# | CUSTOM_PHYSNET_???
# \
# - NIC Switch agent
# | CUSTOM_VNIC_TYPE_DIRECT
# | CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL
# | CUSTOM_VNIC_TYPE_MACVTAP
# \
# \
# - enp129s0f0
# NET_BW_EGR_KILOBIT_PER_SEC: 10000
# NET_BW_IGR_KILOBIT_PER_SEC: 10000
# CUSTOM_VNIC_TYPE_DIRECT
# CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL
# CUSTOM_VNIC_TYPE_MACVTAP
# 'CUSTOM_PHYSNET_???'
#
# For CUSTOM_PHYSNET_??? define the network segment connectivity
# compute0: CUSTOM_PHYSNET_OTHER
# compute1: CUSTOM_PHYSNET_MSN_S1
# compute2: CUSTOM_PHYSNET_MSN_S2
#
# There is a 4th compute that has duplicate network connectivity:
# compute3-br-ex is connected to CUSTOM_PHYSNET_MSN_S1
# compute3-br-ex2 is connected to CUSTOM_PHYSNET_MSN_S2
# compute3-enp129s0f0 is connected to CUSTOM_PHYSNET_MSN_S1
# compute3-enp129s0f1 is connected to CUSTOM_PHYSNET_MSN_S2
# but also compute3 has limited bandwidth capacity
def start_fixture(self):
super(NeutronQoSMultiSegmentFixture, self).start_fixture()
# compute 0 with not connectivity to the multi segment network
compute0 = tb.create_provider(self.context, 'compute0')
os.environ['compute0'] = compute0.uuid
tb.add_inventory(compute0, 'VCPU', 8)
tb.add_inventory(compute0, 'MEMORY_MB', 4096)
tb.add_inventory(compute0, 'DISK_GB', 500)
# OVS agent subtree
compute0_ovs_agent = tb.create_provider(
self.context, 'compute0:Open vSwitch agent', parent=compute0.uuid)
os.environ['compute0:ovs_agent'] = compute0_ovs_agent.uuid
tb.add_inventory(
compute0_ovs_agent, 'NET_PACKET_RATE_KILOPACKET_PER_SEC', 1000)
tb.set_traits(
compute0_ovs_agent,
'CUSTOM_VNIC_TYPE_NORMAL',
)
compute0_br_ex = tb.create_provider(
self.context,
'compute0:Open vSwitch agent:br-ex',
parent=compute0_ovs_agent.uuid
)
os.environ['compute0:br_ex'] = compute0_br_ex.uuid
tb.add_inventory(
compute0_br_ex, 'NET_BW_EGR_KILOBIT_PER_SEC', 5000)
tb.add_inventory(
compute0_br_ex, 'NET_BW_IGR_KILOBIT_PER_SEC', 5000)
tb.set_traits(
compute0_br_ex,
'CUSTOM_VNIC_TYPE_NORMAL',
'CUSTOM_PHYSNET_OTHER',
)
# SRIOV agent subtree
compute0_sriov_agent = tb.create_provider(
self.context, 'compute0:NIC Switch agent', parent=compute0.uuid)
os.environ['compute0:sriov_agent'] = compute0_sriov_agent.uuid
tb.set_traits(
compute0_sriov_agent,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
)
compute0_pf0 = tb.create_provider(
self.context,
'compute0:NIC Switch agent:enp129s0f0',
parent=compute0_sriov_agent.uuid
)
os.environ['compute0:pf0'] = compute0_pf0.uuid
tb.add_inventory(
compute0_pf0, 'NET_BW_EGR_KILOBIT_PER_SEC', 10000)
tb.add_inventory(
compute0_pf0, 'NET_BW_IGR_KILOBIT_PER_SEC', 10000)
tb.set_traits(
compute0_pf0,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
'CUSTOM_PHYSNET_OTHER',
)
# compute 1 with network connectivity to segment 1
compute1 = tb.create_provider(self.context, 'compute1')
os.environ['compute1'] = compute1.uuid
tb.add_inventory(compute1, 'VCPU', 8)
tb.add_inventory(compute1, 'MEMORY_MB', 4096)
tb.add_inventory(compute1, 'DISK_GB', 500)
# OVS agent subtree
compute1_ovs_agent = tb.create_provider(
self.context, 'compute1:Open vSwitch agent', parent=compute1.uuid)
os.environ['compute1:ovs_agent'] = compute1_ovs_agent.uuid
tb.add_inventory(
compute1_ovs_agent, 'NET_PACKET_RATE_KILOPACKET_PER_SEC', 1000)
tb.set_traits(
compute1_ovs_agent,
'CUSTOM_VNIC_TYPE_NORMAL',
)
compute1_br_ex = tb.create_provider(
self.context,
'compute1:Open vSwitch agent:br-ex',
parent=compute1_ovs_agent.uuid
)
os.environ['compute1:br_ex'] = compute1_br_ex.uuid
tb.add_inventory(
compute1_br_ex, 'NET_BW_EGR_KILOBIT_PER_SEC', 5000)
tb.add_inventory(
compute1_br_ex, 'NET_BW_IGR_KILOBIT_PER_SEC', 5000)
tb.set_traits(
compute1_br_ex,
'CUSTOM_VNIC_TYPE_NORMAL',
'CUSTOM_PHYSNET_MSN_S1',
)
# SRIOV agent subtree
compute1_sriov_agent = tb.create_provider(
self.context, 'compute1:NIC Switch agent', parent=compute1.uuid)
os.environ['compute1:sriov_agent'] = compute1_sriov_agent.uuid
tb.set_traits(
compute1_sriov_agent,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
)
compute1_pf0 = tb.create_provider(
self.context,
'compute1:NIC Switch agent:enp129s0f0',
parent=compute1_sriov_agent.uuid
)
os.environ['compute1:pf0'] = compute1_pf0.uuid
tb.add_inventory(
compute1_pf0, 'NET_BW_EGR_KILOBIT_PER_SEC', 10000)
tb.add_inventory(
compute1_pf0, 'NET_BW_IGR_KILOBIT_PER_SEC', 10000)
tb.set_traits(
compute1_pf0,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
'CUSTOM_PHYSNET_MSN_S1',
)
# compute 2 with network connectivity to segment 2
compute2 = tb.create_provider(self.context, 'compute2')
os.environ['compute2'] = compute2.uuid
tb.add_inventory(compute2, 'VCPU', 8)
tb.add_inventory(compute2, 'MEMORY_MB', 4096)
tb.add_inventory(compute2, 'DISK_GB', 500)
# OVS agent subtree
compute2_ovs_agent = tb.create_provider(
self.context, 'compute2:Open vSwitch agent', parent=compute2.uuid)
os.environ['compute2:ovs_agent'] = compute2_ovs_agent.uuid
tb.add_inventory(
compute2_ovs_agent, 'NET_PACKET_RATE_KILOPACKET_PER_SEC', 1000)
tb.set_traits(
compute2_ovs_agent,
'CUSTOM_VNIC_TYPE_NORMAL',
)
compute2_br_ex = tb.create_provider(
self.context,
'compute2:Open vSwitch agent:br-ex',
parent=compute2_ovs_agent.uuid
)
os.environ['compute2:br_ex'] = compute2_br_ex.uuid
tb.add_inventory(
compute2_br_ex, 'NET_BW_EGR_KILOBIT_PER_SEC', 5000)
tb.add_inventory(
compute2_br_ex, 'NET_BW_IGR_KILOBIT_PER_SEC', 5000)
tb.set_traits(
compute2_br_ex,
'CUSTOM_VNIC_TYPE_NORMAL',
'CUSTOM_PHYSNET_MSN_S2',
)
# SRIOV agent subtree
compute2_sriov_agent = tb.create_provider(
self.context, 'compute2:NIC Switch agent', parent=compute2.uuid)
os.environ['compute2:sriov_agent'] = compute2_sriov_agent.uuid
tb.set_traits(
compute2_sriov_agent,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
)
compute2_pf0 = tb.create_provider(
self.context,
'compute2:NIC Switch agent:enp129s0f0',
parent=compute2_sriov_agent.uuid
)
os.environ['compute2:pf0'] = compute2_pf0.uuid
tb.add_inventory(
compute2_pf0, 'NET_BW_EGR_KILOBIT_PER_SEC', 10000)
tb.add_inventory(
compute2_pf0, 'NET_BW_IGR_KILOBIT_PER_SEC', 10000)
tb.set_traits(
compute2_pf0,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
'CUSTOM_PHYSNET_MSN_S2',
)
# compute 3 with network connectivity to both segment 1 and 2
compute3 = tb.create_provider(self.context, 'compute3')
os.environ['compute3'] = compute3.uuid
tb.add_inventory(compute3, 'VCPU', 8)
tb.add_inventory(compute3, 'MEMORY_MB', 4096)
tb.add_inventory(compute3, 'DISK_GB', 500)
# OVS agent subtree
compute3_ovs_agent = tb.create_provider(
self.context, 'compute3:Open vSwitch agent', parent=compute3.uuid)
os.environ['compute3:ovs_agent'] = compute3_ovs_agent.uuid
tb.add_inventory(
compute3_ovs_agent, 'NET_PACKET_RATE_KILOPACKET_PER_SEC', 1000)
tb.set_traits(
compute3_ovs_agent,
'CUSTOM_VNIC_TYPE_NORMAL',
)
compute3_br_ex = tb.create_provider(
self.context,
'compute3:Open vSwitch agent:br-ex',
parent=compute3_ovs_agent.uuid
)
os.environ['compute3:br_ex'] = compute3_br_ex.uuid
tb.add_inventory(
compute3_br_ex, 'NET_BW_EGR_KILOBIT_PER_SEC', 1000)
tb.add_inventory(
compute3_br_ex, 'NET_BW_IGR_KILOBIT_PER_SEC', 1000)
tb.set_traits(
compute3_br_ex,
'CUSTOM_VNIC_TYPE_NORMAL',
'CUSTOM_PHYSNET_MSN_S1',
)
compute3_br_ex2 = tb.create_provider(
self.context,
'compute3:Open vSwitch agent:br-ex2',
parent=compute3_ovs_agent.uuid
)
os.environ['compute3:br_ex2'] = compute3_br_ex2.uuid
tb.add_inventory(
compute3_br_ex2, 'NET_BW_EGR_KILOBIT_PER_SEC', 1000)
tb.add_inventory(
compute3_br_ex2, 'NET_BW_IGR_KILOBIT_PER_SEC', 1000)
tb.set_traits(
compute3_br_ex2,
'CUSTOM_VNIC_TYPE_NORMAL',
'CUSTOM_PHYSNET_MSN_S2',
)
# SRIOV agent subtree
compute3_sriov_agent = tb.create_provider(
self.context, 'compute3:NIC Switch agent', parent=compute3.uuid)
os.environ['compute3:sriov_agent'] = compute2_sriov_agent.uuid
tb.set_traits(
compute3_sriov_agent,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
)
compute3_pf0 = tb.create_provider(
self.context,
'compute3:NIC Switch agent:enp129s0f0',
parent=compute3_sriov_agent.uuid
)
os.environ['compute3:pf0'] = compute3_pf0.uuid
tb.add_inventory(
compute3_pf0, 'NET_BW_EGR_KILOBIT_PER_SEC', 1000)
tb.add_inventory(
compute3_pf0, 'NET_BW_IGR_KILOBIT_PER_SEC', 1000)
tb.set_traits(
compute3_pf0,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
'CUSTOM_PHYSNET_MSN_S1',
)
compute3_pf1 = tb.create_provider(
self.context,
'compute3:NIC Switch agent:enp129s0f1',
parent=compute3_sriov_agent.uuid
)
os.environ['compute3:pf1'] = compute3_pf1.uuid
tb.add_inventory(
compute3_pf1, 'NET_BW_EGR_KILOBIT_PER_SEC', 1000)
tb.add_inventory(
compute3_pf1, 'NET_BW_IGR_KILOBIT_PER_SEC', 1000)
tb.set_traits(
compute3_pf1,
'CUSTOM_VNIC_TYPE_DIRECT',
'CUSTOM_VNIC_TYPE_DIRECT_PHYSICAL',
'CUSTOM_VNIC_TYPE_MACVTAP',
'CUSTOM_PHYSNET_MSN_S2',
)

159
placement/tests/functional/gabbits/allocation-candidates-any-traits-groups.yaml Normal file
View File

@ -0,0 +1,159 @@
fixtures:
- NeutronQoSMultiSegmentFixture
defaults:
request_headers:
x-auth-token: admin
accept: application/json
openstack-api-version: placement latest
tests:
- name: a VM with single port on a non multisegment network
# only compute0 has access to the non-multi-segment network
GET: >-
/allocation_candidates?resources=VCPU:1,MEMORY_MB:1024,DISK_GB:10
&resources-port-normal-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:1000
&required-port-normal-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port-normal-bw=NET_BW_EGR_KILOBIT_PER_SEC:1000,NET_BW_IGR_KILOBIT_PER_SEC:1000
&required-port-normal-bw=CUSTOM_VNIC_TYPE_NORMAL,CUSTOM_PHYSNET_OTHER
&same_subtree=-port-normal-pps,-port-normal-bw
&group_policy=none
status: 200
response_json_paths:
$.allocation_requests.`len`: 1
$.allocation_requests..allocations["$ENVIRON['compute0']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute0']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute0']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute0:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute0:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute0:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 1000
- name: a VM with single port on the multi-segment network
# compute1 compute2 has both access to one segment while compute3 has access
# to two segments so compute1,2 will have one candidate while compute 3 will
# have two
GET: >-
/allocation_candidates?resources=VCPU:1,MEMORY_MB:1024,DISK_GB:10
&resources-port-msn-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:1000
&required-port-msn-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port-msn-bw=NET_BW_EGR_KILOBIT_PER_SEC:1000,NET_BW_IGR_KILOBIT_PER_SEC:1000
&required-port-msn-bw=CUSTOM_VNIC_TYPE_NORMAL
&required-port-msn-bw=in:CUSTOM_PHYSNET_MSN_S1,CUSTOM_PHYSNET_MSN_S2
&same_subtree=-port-msn-pps,-port-msn-bw
&group_policy=none
status: 200
response_json_paths:
$.allocation_requests.`len`: 4
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute1:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute1:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute1:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute2:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute2:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute2:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[VCPU]: [1, 1]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[MEMORY_MB]: [1024, 1024]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[DISK_GB]: [10, 10]
$.allocation_requests..allocations["$ENVIRON['compute3:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: [1000, 1000]
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex2']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 1000
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex2']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 1000
- name: a VM with two ports on the multi-segment network limited bandwidth
# similarly to the single port test compute 1 and compute 2 can offer one
# allocation candidate as both port fits to the one segment of each compute.
# However, compute3 only has enough bandwidth capacity for one port per
# connected network segment. So either we allocate port1-segment1 and
# port2-segment2 OR port1-segment2 and port2-segment1
GET: >-
/allocation_candidates?resources=VCPU:1,MEMORY_MB:1024,DISK_GB:10
&resources-port1-msn-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:100
&required-port1-msn-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port1-msn-bw=NET_BW_EGR_KILOBIT_PER_SEC:1000,NET_BW_IGR_KILOBIT_PER_SEC:1000
&required-port1-msn-bw=CUSTOM_VNIC_TYPE_NORMAL
&required-port1-msn-bw=in:CUSTOM_PHYSNET_MSN_S1,CUSTOM_PHYSNET_MSN_S2
&same_subtree=-port1-msn-pps,-port1-msn-bw
&resources-port2-msn-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:100
&required-port2-msn-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port2-msn-bw=NET_BW_EGR_KILOBIT_PER_SEC:1000,NET_BW_IGR_KILOBIT_PER_SEC:1000
&required-port2-msn-bw=CUSTOM_VNIC_TYPE_NORMAL
&required-port2-msn-bw=in:CUSTOM_PHYSNET_MSN_S1,CUSTOM_PHYSNET_MSN_S2
&same_subtree=-port2-msn-pps,-port2-msn-bw
&group_policy=none
status: 200
response_json_paths:
$.allocation_requests.`len`: 4
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute1:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute1:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 2000
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute2:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute2:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 2000
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[VCPU]: [1, 1]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[MEMORY_MB]: [1024, 1024]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[DISK_GB]: [10, 10]
$.allocation_requests..allocations["$ENVIRON['compute3:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: [200, 200]
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: [1000, 1000]
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex2']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: [1000, 1000]
- name: a VM with two ports on the multi-segment network
# similar test as the previous but the bandwidth request is decreased so
# that compute3 now can fit both ports into one segment. This means compute3
# now has 4 candidates
GET: >-
/allocation_candidates?resources=VCPU:1,MEMORY_MB:1024,DISK_GB:10
&resources-port1-msn-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:100
&required-port1-msn-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port1-msn-bw=NET_BW_EGR_KILOBIT_PER_SEC:100,NET_BW_IGR_KILOBIT_PER_SEC:100
&required-port1-msn-bw=CUSTOM_VNIC_TYPE_NORMAL
&required-port1-msn-bw=in:CUSTOM_PHYSNET_MSN_S1,CUSTOM_PHYSNET_MSN_S2
&same_subtree=-port1-msn-pps,-port1-msn-bw
&resources-port2-msn-pps=NET_PACKET_RATE_KILOPACKET_PER_SEC:100
&required-port2-msn-pps=CUSTOM_VNIC_TYPE_NORMAL
&resources-port2-msn-bw=NET_BW_EGR_KILOBIT_PER_SEC:100,NET_BW_IGR_KILOBIT_PER_SEC:100
&required-port2-msn-bw=CUSTOM_VNIC_TYPE_NORMAL
&required-port2-msn-bw=in:CUSTOM_PHYSNET_MSN_S1,CUSTOM_PHYSNET_MSN_S2
&same_subtree=-port2-msn-pps,-port2-msn-bw
&group_policy=none
status: 200
response_json_paths:
$.allocation_requests.`len`: 6
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute1']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute1:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute1:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute1:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[MEMORY_MB]: 1024
$.allocation_requests..allocations["$ENVIRON['compute2']"].resources[DISK_GB]: 10
$.allocation_requests..allocations["$ENVIRON['compute2:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute2:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute2:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: 200
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[VCPU]: [1, 1, 1, 1]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[MEMORY_MB]: [1024, 1024, 1024, 1024]
$.allocation_requests..allocations["$ENVIRON['compute3']"].resources[DISK_GB]: [10, 10, 10, 10]
$.allocation_requests..allocations["$ENVIRON['compute3:ovs_agent']"].resources[NET_PACKET_RATE_KILOPACKET_PER_SEC]: [200, 200, 200, 200]
# So the 4 candidate from compute3 are
# * both ports allocate from br_ex so br_ex has a consumption of 100 + 100,
# then br_ex2 is not in the candidate (this is why the br_ex2 lists have only 3 items)
# * both ports allocate from br_ex2 then br_ex is not in the candidate (this is why the br_ex lists have only 3 items)
# * port1 allocates 100 from br_ex, port2 allocates 100 from br_ex2
# * port2 allocates 100 from br_ex, port1 allocates 100 from br_ex2
# As the candidates are in random order the right-hand side needs to list all possible permutations
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: /[100, 100, 200]|[100, 200, 100]|[200, 100, 100]/
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: /[100, 100, 200]|[100, 200, 100]|[200, 100, 100]/
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex2']"].resources[NET_BW_IGR_KILOBIT_PER_SEC]: /[100, 100, 200]|[100, 200, 100]|[200, 100, 100]/
$.allocation_requests..allocations["$ENVIRON['compute3:br_ex2']"].resources[NET_BW_EGR_KILOBIT_PER_SEC]: /[100, 100, 200]|[100, 200, 100]|[200, 100, 100]/

28
placement/tests/functional/gabbits/allocation-candidates-any-traits.yaml
View File

@ -46,3 +46,31 @@ tests:
status: 200
response_json_paths:
$.allocation_requests.`len`: 1
- name: get candidates with both OR, AND, and NOT trait queries
# DXVA or TLS would allow all the trees, AVX filters that down to the left
# and the middle but FOO forbids left so middle remains. Middle has access
# to two shared disk provider so the query returns two candidates
GET: /allocation_candidates?required=in:HW_GPU_API_DXVA,HW_NIC_ACCEL_TLS&required=HW_CPU_X86_AVX,!CUSTOM_FOO&resources=VCPU:1,DISK_GB:1
status: 200
response_json_paths:
$.allocation_requests.`len`: 2
$.allocation_requests..allocations["$ENVIRON['CN_MIDDLE']"].resources[VCPU]: [1, 1]
$.allocation_requests..allocations["$ENVIRON['SHR_DISK_1']"].resources[DISK_GB]: 1
$.allocation_requests..allocations["$ENVIRON['SHR_DISK_2']"].resources[DISK_GB]: 1
- name: get candidates with multiple OR queries
# The left tree has neither MMX nor TLS, so it is out. The middle tree has
# TLS and SSD via shr_disk_1 so that is match. The right tree has MMX and SSD
# on the root so that is a match, but it can also get DISK from shr_disk_2
# even if it is not SSD (the SSD trait and the DISK_GB resource are not tight
# together in any way in placement)
GET: /allocation_candidates?required=in:HW_CPU_X86_MMX,HW_NIC_ACCEL_TLS&required=in:CUSTOM_DISK_SSD,CUSTOM_FOO&resources=VCPU:1,DISK_GB:1
status: 200
response_json_paths:
$.allocation_requests.`len`: 3
$.allocation_requests..allocations["$ENVIRON['CN_MIDDLE']"].resources[VCPU]: 1
$.allocation_requests..allocations["$ENVIRON['SHR_DISK_1']"].resources[DISK_GB]: 1
$.allocation_requests..allocations["$ENVIRON['CN_RIGHT']"].resources[VCPU]: [1, 1]
$.allocation_requests..allocations["$ENVIRON['CN_RIGHT']"].resources[DISK_GB]: 1
$.allocation_requests..allocations["$ENVIRON['SHR_DISK_2']"].resources[DISK_GB]: 1

4
placement/tests/functional/gabbits/microversion.yaml
View File

@ -41,13 +41,13 @@ tests:
response_json_paths:
$.errors[0].title: Not Acceptable
- name: latest microversion is 1.38
- name: latest microversion is 1.39
GET: /
request_headers:
openstack-api-version: placement latest
response_headers:
vary: /openstack-api-version/
openstack-api-version: placement 1.38
openstack-api-version: placement 1.39
- name: other accept header bad version
GET: /

18
placement/tests/functional/gabbits/resource-provider-any-traits.yaml
View File

@ -27,3 +27,21 @@ tests:
status: 200
response_json_paths:
$.resource_providers.`len`: 1
- name: list providers with both OR, AND, and NOT trait queries
# DXVA or TLS would allow all the RPs, AVX filters that down to the left and
# the middle but FOO forbids the left so the middle remains
GET: /resource_providers?required=in:HW_GPU_API_DXVA,HW_NIC_ACCEL_TLS&required=HW_CPU_X86_AVX,!CUSTOM_FOO
status: 200
response_json_paths:
$.resource_providers.`len`: 1
$.resource_providers[0].name: cn_middle
- name: have multiple OR queries
# MMX or TLS matches middle and right, SSD or FOO matches left, right and
# shr_disk_1. So only right is a total match.
GET: /resource_providers?required=in:HW_CPU_X86_MMX,HW_NIC_ACCEL_TLS&required=in:CUSTOM_DISK_SSD,CUSTOM_FOO
status: 200
response_json_paths:
$.resource_providers.`len`: 1
$.resource_providers[0].name: cn_right

28
placement/tests/unit/test_util.py
View File

@ -614,10 +614,6 @@ class TestNormalizeTraitsQsParams(testtools.TestCase):
str(ex),
)
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_allow_any_traits_1_39(self):
req = self._get_req('required=in:FOO,BAZ', (1, 39))
@ -626,10 +622,6 @@ class TestNormalizeTraitsQsParams(testtools.TestCase):
self.assertEqual([{'FOO', 'BAZ'}], required)
self.assertEqual(set(), forbidden)
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_repeated_param_1_39(self):
req = self._get_req(
'required=in:T1,T2'
@ -1168,10 +1160,6 @@ class TestParseQsRequestGroups(testtools.TestCase):
"microversion 1.39.",
str(exc))
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_any_traits_1_39(self):
qs = 'resources1=RABBIT:1&required1=in:WHITE,BLACK'
expected = [
@ -1189,10 +1177,6 @@ class TestParseQsRequestGroups(testtools.TestCase):
self.assertRequestGroupsEqual(
expected, self.do_parse(qs, version=(1, 39)))
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_any_traits_repeated(self):
qs = 'resources1=CUSTOM_MAGIC:1&required1=in:T1,T2&required1=T3,!T4'
expected = [
@ -1214,10 +1198,6 @@ class TestParseQsRequestGroups(testtools.TestCase):
self.assertRequestGroupsEqual(
expected, self.do_parse(qs, version=(1, 39)))
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_any_traits_multiple_groups(self):
qs = ('resources=RABBIT:1&required=in:WHITE,BLACK&'
'resources2=CAT:2&required2=in:SILVER,RED&required2=!SPOTTED')
@ -1250,10 +1230,6 @@ class TestParseQsRequestGroups(testtools.TestCase):
self.assertRequestGroupsEqual(
expected, self.do_parse(qs, version=(1, 39)))
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_any_traits_forbidden_conflict(self):
# going against one part of an OR expression is not a conflict as the
# other parts still can match and fulfill the query
@ -1278,10 +1254,6 @@ class TestParseQsRequestGroups(testtools.TestCase):
webob.exc.HTTPBadRequest, self.do_parse, qs, version=(1, 39))
self.assertEqual(expected_message, str(exc))
# TODO(gibi): remove the mock when microversion 1.39 is fully added
@mock.patch(
'placement.microversion.max_version_string',
new=mock.Mock(return_value='1.39'))
def test_stringification(self):
agg1 = uuidsentinel.agg1
agg2 = uuidsentinel.agg2

12
releasenotes/notes/any-traits-support-d3807c27e5a8865c.yaml Normal file
View File

@ -0,0 +1,12 @@
---
features:
- |
Microversion 1.39 adds support for the ``in:`` syntax in the ``required``
query parameter in the ``GET /resource_providers`` API as well as to the
``required`` and ``requiredN`` query params of the
``GET /allocation_candidates`` API. Also adds support for repeating the
``required`` and ``requiredN`` parameters in the respective APIs. So::
required=in:T3,T4&required=T1,!T2
is supported and it means T1 and not T2 and (T3 or T4).