1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
|
# Copyright (c) 2018 Intel.
# All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""
Track resources like FPGA GPU and QAT for a host. Provides the
conductor with useful information about availability through the accelerator
model.
"""
from oslo_log import log as logging
from oslo_messaging.rpc.client import RemoteError
from oslo_utils import uuidutils
from cyborg.accelerator.drivers.fpga.base import FPGADriver
from cyborg.accelerator.drivers.gpu.base import GPUDriver
from cyborg.common import utils
from cyborg import objects
LOG = logging.getLogger(__name__)
AGENT_RESOURCE_SEMAPHORE = "agent_resources"
DEPLOYABLE_VERSION = "1.0"
# need to change the driver field name
DEPLOYABLE_HOST_MAPS = {"assignable": "assignable",
"pcie_address": "devices",
"board": "product_id",
"type": "function",
"vendor": "vendor_id",
"name": "name"}
class ResourceTracker(object):
"""Agent helper class for keeping track of resource usage when hardware
Accelerator resources updated. Update the Deployable DB through conductor.
"""
def __init__(self, host, cond_api):
# FIXME (Shaohe) local cache for Accelerator.
# Will fix it in next release.
self.fpgas = None
self.host = host
self.conductor_api = cond_api
self.fpga_driver = FPGADriver()
self.gpu_driver = GPUDriver()
@utils.synchronized(AGENT_RESOURCE_SEMAPHORE)
def claim(self, context):
pass
def _fpga_compare_and_update(self, host_dev, acclerator):
need_updated = False
for k, v in DEPLOYABLE_HOST_MAPS.items():
if acclerator[k] != host_dev[v]:
need_updated = True
acclerator[k] = host_dev[v]
return need_updated
def _gen_deployable_from_host_dev(self, host_dev):
dep = {}
for k, v in DEPLOYABLE_HOST_MAPS.items():
dep[k] = host_dev[v]
dep["host"] = self.host
dep["version"] = DEPLOYABLE_VERSION
dep["availability"] = "free"
dep["uuid"] = uuidutils.generate_uuid()
return dep
@utils.synchronized(AGENT_RESOURCE_SEMAPHORE)
def update_usage(self, context):
"""Update the resource usage and stats after a change in an
instance
"""
def create_deployable(fpgas, bdf, parent_uuid=None):
fpga = fpgas[bdf]
dep = self._gen_deployable_from_host_dev(fpga)
# if parent_uuid:
dep["parent_uuid"] = parent_uuid
obj_dep = objects.Deployable(context, **dep)
new_dep = self.conductor_api.deployable_create(context, obj_dep)
return new_dep
self.update_gpu_usage(context)
# NOTE(Shaohe Feng) need more agreement on how to keep consistency.
fpgas = self._get_fpga_devices()
bdfs = set(fpgas.keys())
deployables = self.conductor_api.deployable_get_by_host(
context, self.host)
# NOTE(Shaohe Feng) when no "pcie_address" in deployable?
accls = dict([(v["pcie_address"], v) for v in deployables
if v["type"] == "FPGA"])
accl_bdfs = set(accls.keys())
# Firstly update
for mutual in accl_bdfs & bdfs:
accl = accls[mutual]
if self._fpga_compare_and_update(fpgas[mutual], accl):
try:
self.conductor_api.deployable_update(context, accl)
except RemoteError as e:
LOG.error(e)
# Add
new = bdfs - accl_bdfs
new_pf = set([n for n in new if fpgas[n]["function"] == "pf"])
for n in new_pf:
new_dep = create_deployable(fpgas, n)
accls[n] = new_dep
sub_vf = set()
if "regions" in n:
sub_vf = set([sub["devices"] for sub in fpgas[n]["regions"]])
for vf in sub_vf & new:
new_dep = create_deployable(fpgas, vf, new_dep["uuid"])
accls[vf] = new_dep
new.remove(vf)
for n in new - new_pf:
p_bdf = fpgas[n]["parent_devices"]
p_accl = accls[p_bdf]
p_uuid = p_accl["uuid"]
new_dep = create_deployable(fpgas, n, p_uuid)
# Delete
for obsolete in accl_bdfs - bdfs:
try:
self.conductor_api.deployable_delete(context, accls[obsolete])
except RemoteError as e:
LOG.error(e)
del accls[obsolete]
def _get_fpga_devices(self):
def form_dict(devices, fpgas):
for v in devices:
fpgas[v["devices"]] = v
if "regions" in v:
form_dict(v["regions"], fpgas)
fpgas = {}
vendors = self.fpga_driver.discover_vendors()
for v in vendors:
driver = self.fpga_driver.create(v)
form_dict(driver.discover(), fpgas)
return fpgas
def update_gpu_usage(self, context):
"""Update the gpu resource usage and stats after a change in an
instance, for the original update_usage specified update fpga, define a
new func update gpu here.
"""
def create_deployable(gpus, bdf, parent_uuid=None):
gpu = gpus[bdf]
dep = self._gen_deployable_from_host_dev(gpu)
# if parent_uuid:
dep["parent_uuid"] = parent_uuid
obj_dep = objects.Deployable(context, **dep)
new_dep = self.conductor_api.deployable_create(context, obj_dep)
return new_dep
gpus = self._get_gpu_devices()
deployables = self.conductor_api.deployable_get_by_host(
context, self.host)
accls = dict([(v["pcie_address"], v) for v in deployables
if v["type"] == "GPU"])
all_gpus = dict([(v["devices"], v) for v in gpus])
# Add
new = set(all_gpus.keys()) - set(accls.keys())
new_gpus = [all_gpus[n] for n in new]
for n in new_gpus:
dep = self._gen_deployable_from_host_dev(n)
# if parent_uuid:
dep["parent_uuid"] = None
obj_dep = objects.Deployable(context, **dep)
self.conductor_api.deployable_create(context, obj_dep)
# Delete
not_exists = set(accls.keys()) - set(all_gpus.keys())
for obsolete in not_exists:
try:
self.conductor_api.deployable_delete(context, accls[obsolete])
except RemoteError as e:
LOG.error(e)
del accls[obsolete]
def _get_gpu_devices(self):
gpus = []
vendors = self.gpu_driver.discover_vendors()
for v in vendors:
driver = self.gpu_driver.create(v)
if driver:
gpus.extend(driver.discover())
return gpus
|
