@@ -175,6 +175,81 @@ spec:
175
175
x-kubernetes-validations :
176
176
- rule : self.matches(r'^([a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*/)?([a-z0-9]([-a-z0-9]*[a-z0-9])?)$') && self.size() <= 317
177
177
x-kubernetes-list-type : atomic
178
+ <<<<<<< HEAD
179
+ =======
180
+ kueueGangSchedulingPolicy :
181
+ description : |-
182
+ kueueGangSchedulingPolicy controls how Kueue admits workloads.
183
+ Gang Scheduling is the act of all or nothing scheduling.
184
+ Kueue provides this ability.
185
+ This field is optional.
186
+ type : object
187
+ properties :
188
+ byWorkload :
189
+ description : |-
190
+ byWorkload controls how admission is done.
191
+ When admission is set to Sequential, only pods from the currently processing workload will be admitted.
192
+ Once all pods from the current workload are admitted, and ready, Kueue will process the next workload.
193
+ Sequential processing may slow down admission when the cluster has sufficient capacity for multiple workloads,
194
+ but provides a higher guarantee of workloads scheduling all pods together successfully.
195
+ When set to Parallel, pods from any workload will be admitted at any time.
196
+ This may lead to a deadlock where workloads are in contention for cluster capacity and
197
+ pods from another workload having successfully scheduled prevent pods from the current workload scheduling.
198
+ type : string
199
+ enum :
200
+ - Parallel
201
+ - Sequential
202
+ x-kubernetes-validations :
203
+ - rule : self.policy==ByWorkload
204
+ message : byWorkload is only valid if policy equals ByWorkload
205
+ policy :
206
+ description : |-
207
+ policy allows for changing the kinds of gang scheduling Kueue does.
208
+ This is an optional field.
209
+ The allowed values are ByWorkload and Disabled.
210
+ The default value will be Disabled.
211
+ ByWorkload allows for configuration how admission is performed
212
+ for Kueue.
213
+ type : string
214
+ enum :
215
+ - ByWorkload
216
+ - Disabled
217
+ preemption :
218
+ description : |-
219
+ preemption is the process of evicting one or more admitted Workloads to accommodate another Workload.
220
+ Kueue has classical preemption and preemption via fair sharing.
221
+ type : object
222
+ properties :
223
+ preemptionStrategy :
224
+ description : |-
225
+ preemptionStrategy are the types of preemption kueue allows.
226
+ Kueue has two types of preemption: classical and fair sharing.
227
+ type : string
228
+ enum :
229
+ - Classical
230
+ - FairSharing
231
+ queueLabelPolicy :
232
+ description : |-
233
+ queueLabelPolicy controls how kueue manages workloads
234
+ The default behavior of Kueue will manage workloads that have a queue-name label.
235
+ This field is optional.
236
+ type : object
237
+ properties :
238
+ queueLabelPolicy :
239
+ description : |-
240
+ queueLabelPolicy controls whether or not Kueue reconciles
241
+ jobs that don't set the label kueue.x-k8s.io/queue-name.
242
+ The allowed values are QueueNameRequired and QueueNameOptional.
243
+ If set to QueueNameRequired, then those jobs will be suspended and never started unless
244
+ they are assigned a queue and eventually admitted. This also applies to
245
+ jobs created before starting the kueue controller.
246
+ Defaults to QueueNameRequired; therefore, those jobs are not managed and if they are created
247
+ unsuspended, they will start immediately.
248
+ type : string
249
+ enum :
250
+ - QueueNameRequired
251
+ - QueueNameOptional
252
+ >>>>>>> 22bc930 (add FairSharing, QueueLabelPolicy and GangScheduling)
178
253
logLevel :
179
254
description : |-
180
255
logLevel is an intent based logging for an overall component. It does not give fine grained control, but it is a
0 commit comments