Merge pull request #2539 from consideRatio/pr/awi-ciroh-event

awi-ciroh: last minute fixes for event 16-18 May - node sharing adopted to avoid PD quota, legacy profile list retained

config/clusters/awi-ciroh/common.values.yaml

@@ -54,42 +54,67 @@ basehub:
     singleuser:
       image:
         # Image build repo: https://github.com/2i2c-org/awi-ciroh-image
+        #
+        # NOTE: The configurator is used in this cluster, so this is stale
+        #       configuration.
+        #
         name: "quay.io/2i2c/awi-ciroh-image"
-        tag: "63ecd92f8d84"
+        tag: "584293e50d4c"
       profileList:
         # The mem-guarantees are here so k8s doesn't schedule other pods
         # on these nodes. They need to be just under total allocatable
         # RAM on a node, not total node capacity. Values calculated using
         # https://learnk8s.io/kubernetes-instance-calculator
+        #
+        # FIXME: These are changed to a temporary node sharing setup based on
+        #        the legacy choices to help us pre-warm capacity independent on
+        #        the choices users end up making and avoiding running into
+        #        persistent disk quotas.
+        #
+        #        Change PR: https://github.com/2i2c-org/infrastructure/pull/2539
+        #        Related event: https://github.com/2i2c-org/infrastructure/issues/2520
+        #
+        #        This is an interim setup, trying to balance the experience of
+        #        the previous 1:1 user:node setup with a node sharing setup. It
+        #        is not meant to be retained long term!
+        #
         - display_name: "Small"
           description: 5GB RAM, 2 CPUs
           default: true
           kubespawner_override:
             mem_limit: 7G
-            mem_guarantee: 4.5G
+            mem_guarantee: 5G
+            cpu_limit: 2
+            cpu_guarantee: 0.938
             node_selector:
-              node.kubernetes.io/instance-type: n1-standard-2
+              node.kubernetes.io/instance-type: n2-highmem-16
         - display_name: Medium
           description: 11GB RAM, 4 CPUs
           kubespawner_override:
             mem_limit: 15G
             mem_guarantee: 11G
+            cpu_limit: 4
+            cpu_guarantee: 1.875
             node_selector:
-              node.kubernetes.io/instance-type: n1-standard-4
+              node.kubernetes.io/instance-type: n2-highmem-16
         - display_name: Large
           description: 24GB RAM, 8 CPUs
           kubespawner_override:
             mem_limit: 30G
             mem_guarantee: 24G
+            cpu_limit: 8
+            cpu_guarantee: 3.75
             node_selector:
-              node.kubernetes.io/instance-type: n1-standard-8
+              node.kubernetes.io/instance-type: n2-highmem-16
         - display_name: Huge
           description: 52GB RAM, 16 CPUs
           kubespawner_override:
             mem_limit: 60G
             mem_guarantee: 52G
+            cpu_limit: 16
+            cpu_guarantee: 7.5
             node_selector:
-              node.kubernetes.io/instance-type: n1-standard-16
+              node.kubernetes.io/instance-type: n2-highmem-16
 dask-gateway:
   gateway:
     backend:

terraform/gcp/projects/awi-ciroh.tfvars

@@ -68,6 +68,24 @@ notebook_nodes = {
       count: 0
     }
   },
+  # added stressfully before an event where we ran out of ssd quota, see
+  # https://github.com/2i2c-org/infrastructure/pull/2539 and the linked
+  # event https://github.com/2i2c-org/infrastructure/issues/2520.
+  #
+  # FIXME: make this cluster have a node sharing setup like in the
+  #        basehub/daskhub template.
+  #
+  "highmem-medium" : {
+    min : 10,
+    max : 100,
+    machine_type : "n2-highmem-16",
+    labels: {},
+    gpu: {
+      enabled: false,
+      type: "",
+      count: 0
+    }
+  },
 }
 
 dask_nodes = {