update tests (#111)

Signed-off-by: Dmitry Shmulevich <[email protected]>

Author: dmitsh

docs/examples/kueue/kueue.md

@@ -3,7 +3,7 @@
 Install `kueue` by following these [instructions](https://kueue.sigs.k8s.io/docs/installation/):
 
 ```bash
-KUEUE_VERSION=v0.8.0
+KUEUE_VERSION=v0.9.0
 kubectl apply --server-side -f https://github.com/kubernetes-sigs/kueue/releases/download/${KUEUE_VERSION}/manifests.yaml
 
 kubectl apply -f charts/overrides/kueue/priority.yaml

resources/benchmarks/README.md

@@ -36,7 +36,7 @@ To run the benchmark test for Run:ai
 
 ## Scaling Benchmark Test
 
-The scaling benchmark workflow operates on 500 virtual GPU nodes with tho workflows. The first [workflow](scaling/workflows/run-test-multi.yaml) submits is a job with 500 replicas, the second [workflow](scaling/workflows/run-test-single.yaml) submits a batch of 500 single-node jobs.
+The scaling benchmark workflow operates on 700 virtual GPU nodes with tho workflows. The first [workflow](scaling/workflows/run-test-multi.yaml) submits is a job with 700 replicas, the second [workflow](scaling/workflows/run-test-single.yaml) submits a batch of 700 single-node jobs.
 
 ### Example
 

resources/benchmarks/nwtopo/templates/runai/mpijob.yaml

@@ -43,15 +43,6 @@ spec:
           affinity:
             podAffinity:
               preferredDuringSchedulingIgnoredDuringExecution:
-                - weight: 20
-                  podAffinityTerm:
-                    labelSelector:
-                      matchExpressions:
-                        - key: app
-                          operator: In
-                          values:
-                            - {{._NAME_}}
-                    topologyKey: net-layer-3
                 - weight: 70
                   podAffinityTerm:
                     labelSelector:

resources/benchmarks/scaling/workflows/config-nodes.yaml

@@ -13,14 +13,14 @@
 # limitations under the License.
 
 name: config-nodes
-description: create 500 virtual GPU nodes
+description: create 700 virtual GPU nodes
 tasks:
 - id: configure
   type: Configure
   params:
     nodes:
     - type: dgxa100.80g
-      count: 500
+      count: 700
       labels:
         nvidia.com/gpu.count: "8"
     timeout: 5m

resources/benchmarks/scaling/workflows/config-yunikorn.yaml

@@ -40,5 +40,5 @@ tasks:
                   submitacl: '*'
                   resources:
                     max:
-                      {memory: 360Gi, vcore: 50000m, nvidia.com/gpu: 4000}
+                      {memory: 360Gi, vcore: 70000m, nvidia.com/gpu: 5600}
     timeout: 1m

resources/benchmarks/scaling/workflows/run-test-multi.yaml

@@ -13,13 +13,13 @@
 # limitations under the License.
 
 name: test-scaling-multi-node-job
-description: deploy a 500-replicas job
+description: deploy a 700-replicas job
 tasks:
 - id: job
   type: SubmitObj
   params:
     refTaskId: register
     count: 1
     params:
-      replicas: 500
-      ttl: 2m
+      replicas: 700
+      ttl: 5m

resources/benchmarks/scaling/workflows/run-test-single.yaml

@@ -13,13 +13,13 @@
 # limitations under the License.
 
 name: test-scaling-single-node-jobs
-description: deploy 500 single-replica jobs
+description: deploy 700 single-replica jobs
 tasks:
 - id: job
   type: SubmitObj
   params:
     refTaskId: register
-    count: 500
+    count: 700
     params:
       replicas: 1
-      ttl: 2m
+      ttl: 5m

resources/benchmarks/scaling/workflows/runai-test-multi.yaml

@@ -13,13 +13,13 @@
 # limitations under the License.
 
 name: test-scaling
-description: deploy a 500-replicas job
+description: deploy a 700-replicas job
 tasks:
 - id: job
   type: SubmitObj
   params:
     refTaskId: register-mpi
     count: 1
     params:
-     workers: 499
-     ttl: 2m
+     workers: 699
+     ttl: 5m

resources/benchmarks/scaling/workflows/runai-test-single.yaml

@@ -13,12 +13,12 @@
 # limitations under the License.
 
 name: test-scaling
-description: deploy 500 single-replica jobs
+description: deploy 700 single-replica jobs
 tasks:
 - id: job
   type: SubmitObj
   params:
     refTaskId: register-trainingworkload
-    count: 500
+    count: 700
     params:
-      ttl: 2m
+      ttl: 5m

scripts/env.sh

@@ -131,7 +131,7 @@ function deploy_jobset() {
 }
 
 # https://github.com/kubernetes-sigs/kueue
-KUEUE_VERSION=v0.8.1
+KUEUE_VERSION=v0.9.0
 
 function deploy_kueue() {
   printGreen Deploying kueue