Enable enableClusterQueueResources metrics (#50)

project-codeflare · Sep 6, 2024 · 65b1c6c · 65b1c6c
1 parent 144a104
commit 65b1c6c
Show file tree

Hide file tree

Showing 12 changed files with 13 additions and 6 deletions.
diff --git a/setup.RHOAI-v2.10/CLUSTER-SETUP.md b/setup.RHOAI-v2.10/CLUSTER-SETUP.md
@@ -79,6 +79,7 @@ AI configuration as follows:
   - `batch/job` integration is disabled,
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - Codeflare operator:
   - the AppWrapper controller is enabled and configured as follows:
     - `userRBACAdmissionCheck` is disabled,

diff --git a/setup.RHOAI-v2.10/mlbatch-subscription.yaml b/setup.RHOAI-v2.10/mlbatch-subscription.yaml
@@ -139,7 +139,7 @@ data:
       healthProbeBindAddress: :8081
     metrics:
       bindAddress: :8080
-    # enableClusterQueueResources: true
+      enableClusterQueueResources: true
     webhook:
       port: 9443
     leaderElection:

diff --git a/setup.RHOAI-v2.11/CLUSTER-SETUP.md b/setup.RHOAI-v2.11/CLUSTER-SETUP.md
@@ -79,6 +79,7 @@ AI configuration as follows:
   - `batch/job` integration is disabled,
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - Codeflare operator:
   - the AppWrapper controller is enabled and configured as follows:
     - `userRBACAdmissionCheck` is disabled,

diff --git a/setup.RHOAI-v2.11/mlbatch-subscription.yaml b/setup.RHOAI-v2.11/mlbatch-subscription.yaml
@@ -139,7 +139,7 @@ data:
       healthProbeBindAddress: :8081
     metrics:
       bindAddress: :8080
-    # enableClusterQueueResources: true
+      enableClusterQueueResources: true
     webhook:
       port: 9443
     leaderElection:

diff --git a/setup.RHOAI-v2.11/mlbatch-upgrade-configmaps.yaml b/setup.RHOAI-v2.11/mlbatch-upgrade-configmaps.yaml
@@ -27,7 +27,7 @@ data:
       healthProbeBindAddress: :8081
     metrics:
       bindAddress: :8080
-    # enableClusterQueueResources: true
+      enableClusterQueueResources: true
     webhook:
       port: 9443
     leaderElection:

diff --git a/setup.RHOAI-v2.12/CLUSTER-SETUP.md b/setup.RHOAI-v2.12/CLUSTER-SETUP.md
@@ -79,6 +79,7 @@ AI configuration as follows:
   - `batch/job` integration is disabled,
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - Codeflare operator:
   - the AppWrapper controller is enabled and configured as follows:
     - `userRBACAdmissionCheck` is disabled,

diff --git a/setup.RHOAI-v2.12/mlbatch-subscription.yaml b/setup.RHOAI-v2.12/mlbatch-subscription.yaml
@@ -152,7 +152,7 @@ data:
       healthProbeBindAddress: :8081
     metrics:
       bindAddress: :8080
-    # enableClusterQueueResources: true
+      enableClusterQueueResources: true
     webhook:
       port: 9443
     leaderElection:

diff --git a/setup.k8s-v1.25/CLUSTER-SETUP.md b/setup.k8s-v1.25/CLUSTER-SETUP.md
@@ -79,6 +79,7 @@ operators as follows:
   - `batch/job` integration is disabled,
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - AppWrapper operator:
   - `userRBACAdmissionCheck` is disabled,
   - `schedulerName` is set to `scheduler-plugins-scheduler`,

diff --git a/setup.k8s-v1.25/kueue/controller_manager_config.yaml b/setup.k8s-v1.25/kueue/controller_manager_config.yaml
@@ -4,7 +4,7 @@ health:
   healthProbeBindAddress: :8081
 metrics:
   bindAddress: :8080
-# enableClusterQueueResources: true
+  enableClusterQueueResources: true
 webhook:
   port: 9443
 leaderElection:

diff --git a/setup.k8s-v1.30/CLUSTER-SETUP.md b/setup.k8s-v1.30/CLUSTER-SETUP.md
@@ -77,6 +77,7 @@ operators as follows:
 - Kueue:
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - AppWrapper operator:
   - `userRBACAdmissionCheck` is disabled,
   - `schedulerName` is set to `scheduler-plugins-scheduler`,

diff --git a/setup.k8s-v1.30/kueue/controller_manager_config.yaml b/setup.k8s-v1.30/kueue/controller_manager_config.yaml
@@ -4,7 +4,7 @@ health:
   healthProbeBindAddress: :8081
 metrics:
   bindAddress: :8080
-# enableClusterQueueResources: true
+  enableClusterQueueResources: true
 webhook:
   port: 9443
 leaderElection:

diff --git a/setup.tmpl/CLUSTER-SETUP.md.tmpl b/setup.tmpl/CLUSTER-SETUP.md.tmpl
@@ -105,6 +105,7 @@ AI configuration as follows:
   - `batch/job` integration is disabled,
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - Codeflare operator:
   - the AppWrapper controller is enabled and configured as follows:
     - `userRBACAdmissionCheck` is disabled,
@@ -170,6 +171,7 @@ operators as follows:
 {{- end }}
   - `waitForPodsReady` is disabled,
   - `LendingLimit` feature gate is enabled,
+  - `enableClusterQueueResources` metrics is enabled,
 - AppWrapper operator:
   - `userRBACAdmissionCheck` is disabled,
   - `schedulerName` is set to `scheduler-plugins-scheduler`,