Revert "CHS Integration for A3 Ultra"

Author: ishitachail

examples/gke-a3-megagpu/gke-a3-megagpu-deployment.yaml

@@ -46,5 +46,5 @@ vars:
   # can be inputted as <reservation-name>/reservationBlocks/<reservation-block-name>
   reservation: RESERVATION_NAME
 
-  enable_periodic_health_checks: # Make this true to run CHS (healthchecks)
-  health_check_schedule:  # Run the health check at 12:00 AM (midnight) every Sunday
+  enable_periodic_health_checks: false # Make this true to run CHS (healthchecks)
+  health_check_schedule: "0 0 * * 0" # Run the health check at 12:00 AM (midnight) every Sunday

examples/gke-a3-ultragpu/chs-cronjob.yaml.tftpl

@@ -36,5 +36,3 @@ vars:
   # disk for each node of the system node pool.
   a3ultra_node_pool_disk_size_gb: A3ULTRA_NODE_POOL_DISK_SIZE_GB # the size of
   # disk for each node.
-  enable_periodic_health_checks: # Make this true to run CHS (healthchecks)
-  health_check_schedule: # Run the health check at 12:00 AM (midnight) every Sunday

examples/gke-a3-ultragpu/chs-permissions.yaml.tftpl

@@ -35,15 +35,6 @@ vars:
   accelerator_type: nvidia-h200-141gb
   version_prefix: "1.31."
 
-  enable_periodic_health_checks: false # Make this true to run CHS (healthchecks)
-  health_check_schedule: "0 0 * * 0" # Run the health check at 12:00 AM (midnight) every Sunday
-
-  permissions_file_staged_path: $(ghpc_stage("./chs-permissions.yaml.tftpl"))
-  chs_output_bucket_name: chs-result
-  chs_pvc_claim_name: chs-output-pvc
-  chs_cronjob_rendered_path: $(ghpc_stage("./chs-cronjob.yaml.tftpl"))
-  chs_pvc_rendered_path: $(ghpc_stage("./chs-pvc.yaml.tftpl"))
-
 deployment_groups:
 - group: primary
   modules:
@@ -125,7 +116,6 @@ deployment_groups:
       - stackdriver.resourceMetadata.writer
       - storage.objectAdmin
       - artifactregistry.reader
-      - container.admin
 
   - id: training_bucket
     source: community/modules/file-system/cloud-storage-bucket
@@ -224,27 +214,6 @@ deployment_groups:
     source: modules/management/kubectl-apply
     use: [a3-ultragpu-cluster]
     settings:
-      apply_manifests:
-      - source: $(vars.permissions_file_staged_path)
-        template_vars:
-          project_id: $(vars.project_id)
-      - source: $(vars.chs_pvc_rendered_path)
-        enable: $(vars.enable_periodic_health_checks)
-        template_vars:
-          pvc_name: $(vars.chs_pvc_claim_name)
-          access_mode: ReadWriteOnce
-          capacity: 1Gi
-          storage_class_name: standard-rwo
-      - source: $(vars.chs_cronjob_rendered_path)
-        enable: $(vars.enable_periodic_health_checks)
-        template_vars:
-          project_id: $(vars.project_id)
-          deployment_name: $(vars.deployment_name)
-          region: $(vars.region)
-          machine_type: a3-ultragpu-8g
-          gcs_bucket: $(vars.chs_output_bucket_name)
-          gcs_pvc: $(vars.chs_pvc_claim_name)
-          cronjob_schedule: $(vars.health_check_schedule)
       kueue:
         install: true
         config_path: $(vars.kueue_configuration_path)