Merge pull request #3131 from consideRatio/pr/2i2c-pilot-hubs-update-…

…core-node-pool

gcp: k8s version updates, transitions to pd-balanced disks, towards n2- nodes

config/clusters/linked-earth/common.values.yaml

@@ -109,7 +109,7 @@ basehub:
             cpu_limit: null
             mem_limit: null
             node_selector:
-              node.kubernetes.io/instance-type: e2-highmem-4
+              node.kubernetes.io/instance-type: n2-highmem-4
         - display_name: "Medium: up to 16 CPU / 128 GB RAM"
           description: *profile_list_description
           slug: medium
@@ -165,7 +165,7 @@ basehub:
             cpu_limit: null
             mem_limit: null
             node_selector:
-              node.kubernetes.io/instance-type: e2-highmem-16
+              node.kubernetes.io/instance-type: n2-highmem-16
 dask-gateway:
   gateway:
     backend:

config/clusters/meom-ige/cluster.yaml

@@ -1,5 +1,5 @@
 name: meom-ige
-provider: gcp # https://console.cloud.google.com/kubernetes/clusters/details/us-central1-b/pangeo-hubs-cluster/nodes?project=columbia
+provider: gcp # https://console.cloud.google.com/kubernetes/clusters/details/us-central1-b/meom-ige-cluster/details?authuser=3&project=meom-ige-cnrs
 gcp:
   key: enc-deployer-credentials.secret.json
   project: meom-ige-cnrs

terraform/gcp/cluster.tf

@@ -185,7 +185,6 @@ resource "google_container_node_pool" "core" {
   location = google_container_cluster.cluster.location
   version  = var.k8s_versions.core_nodes_version
 
-
   initial_node_count = 1
   autoscaling {
     min_node_count = 1
@@ -213,6 +212,12 @@ resource "google_container_node_pool" "core" {
 
 
   node_config {
+    # Balanced disks are much faster than standard disks, and much cheaper
+    # than SSD disks. It contributes heavily to how fast new nodes spin up,
+    # as images being pulled takes up a lot of new node spin up time.
+    # Faster disks provide faster image pulls!
+    disk_type = "pd-balanced"
+
     labels = {
       "hub.jupyter.org/node-purpose" = "core",
       "k8s.dask.org/node-purpose"    = "core"

terraform/gcp/projects/2i2c-uk.tfvars

@@ -1,10 +1,17 @@
 prefix     = "two-eye-two-see-uk"
 project_id = "two-eye-two-see-uk"
 
-zone   = "europe-west2-b"
-region = "europe-west2"
+zone             = "europe-west2-b"
+region           = "europe-west2"
+regional_cluster = true
 
-core_node_machine_type = "n1-highmem-4"
+k8s_versions = {
+  min_master_version : "1.27.4-gke.900",
+  core_nodes_version : "1.27.4-gke.900",
+  notebook_nodes_version : "1.27.4-gke.900",
+}
+
+core_node_machine_type = "n2-highmem-4"
 enable_network_policy  = true
 
 # Setup a filestore for in-cluster NFS
@@ -15,7 +22,7 @@ notebook_nodes = {
   "user" : {
     min : 0,
     max : 20,
-    machine_type : "n1-highmem-4"
+    machine_type : "n2-highmem-4"
   },
 }
 

terraform/gcp/projects/callysto.tfvars

@@ -1,13 +1,14 @@
 prefix     = "callysto"
 project_id = "callysto-202316"
 
-zone   = "northamerica-northeast1-b"
-region = "northamerica-northeast1"
+zone             = "northamerica-northeast1-b"
+region           = "northamerica-northeast1"
+regional_cluster = true
 
 k8s_versions = {
-  min_master_version : "1.25.6-gke.1000",
-  core_nodes_version : "1.25.6-gke.1000",
-  notebook_nodes_version : "1.25.6-gke.1000",
+  min_master_version : "1.27.4-gke.900",
+  core_nodes_version : "1.27.4-gke.900",
+  notebook_nodes_version : "1.27.4-gke.900",
 }
 
 core_node_machine_type = "n2-highmem-2"

terraform/gcp/projects/cloudbank.tfvars

@@ -1,19 +1,26 @@
 prefix     = "cb"
 project_id = "cb-1003-1696"
 
-zone   = "us-central1-b"
-region = "us-central1"
+zone             = "us-central1-b"
+region           = "us-central1"
+regional_cluster = false
+
+k8s_versions = {
+  min_master_version : "1.26.5-gke.2100",
+  core_nodes_version : "1.26.5-gke.2100",
+  notebook_nodes_version : "1.26.4-gke.1400",
+}
 
+# FIXME: We have a temporary core node pool setup with n2-highmem-4 and
+#        pd-balanced. This node pool still has standard though, but has been
+#        cordoned.
+#
 core_node_machine_type = "n1-highmem-4"
+enable_network_policy  = true
 
 enable_filestore      = true
 filestore_capacity_gb = 1024
 
-# Multi-tenant cluster, network policy is required to enforce separation between hubs
-enable_network_policy = true
-
-regional_cluster = false
-
 notebook_nodes = {
   "user" : {
     min : 0,

terraform/gcp/projects/linked-earth.tfvars

@@ -1,11 +1,20 @@
-prefix                 = "linked-earth"
-project_id             = "linked-earth-hubs"
-zone                   = "us-central1-c"
-region                 = "us-central1"
-core_node_machine_type = "e2-highmem-4"
+prefix     = "linked-earth"
+project_id = "linked-earth-hubs"
+
+zone             = "us-central1-c"
+region           = "us-central1"
+regional_cluster = true
+
+k8s_versions = {
+  min_master_version : "1.27.4-gke.900",
+  core_nodes_version : "1.27.4-gke.900",
+  notebook_nodes_version : "1.27.4-gke.900",
+  dask_nodes_version : "1.27.4-gke.900",
+}
+
+core_node_machine_type = "n2-highmem-4"
 enable_network_policy  = true
 
-# Setup a filestore for in-cluster NFS
 enable_filestore      = true
 filestore_capacity_gb = 1024
 
@@ -23,12 +32,12 @@ notebook_nodes = {
   "small" : {
     min : 0,
     max : 100,
-    machine_type : "e2-highmem-4"
+    machine_type : "n2-highmem-4"
   },
   "medium" : {
     min : 0,
     max : 100,
-    machine_type : "e2-highmem-16"
+    machine_type : "n2-highmem-16"
   },
 }
 

terraform/gcp/projects/m2lines.tfvars

@@ -1,14 +1,22 @@
-prefix                 = "m2lines"
-project_id             = "m2lines-hub"
-core_node_machine_type = "n1-highmem-4"
-
-enable_network_policy = true
+prefix     = "m2lines"
+project_id = "m2lines-hub"
 
 # GPUs not available in us-central1-b
 zone             = "us-central1-c"
 region           = "us-central1"
 regional_cluster = true
 
+k8s_versions = {
+  min_master_version : "1.27.4-gke.900",
+  core_nodes_version : "1.27.4-gke.900",
+  notebook_nodes_version : "1.27.4-gke.900",
+  dask_nodes_version : "1.27.4-gke.900",
+}
+
+core_node_machine_type = "n2-highmem-4"
+enable_network_policy  = true
+
+
 # Setup a filestore for in-cluster NFS
 enable_filestore      = true
 filestore_capacity_gb = 2048

terraform/gcp/projects/meom-ige.tfvars

@@ -1,15 +1,18 @@
 prefix     = "meom-ige"
 project_id = "meom-ige-cnrs"
 
-zone   = "us-central1-b"
-region = "us-central1"
+zone             = "us-central1-b"
+region           = "us-central1"
+regional_cluster = false
 
-core_node_machine_type = "n1-highmem-2"
+k8s_versions = {
+  min_master_version : "1.27.4-gke.900",
+  core_nodes_version : "1.27.4-gke.900",
+  notebook_nodes_version : "1.27.4-gke.900",
+}
 
-# Single-tenant cluster, network policy not needed
-enable_network_policy = false
-
-regional_cluster = false
+core_node_machine_type = "n2-highmem-4"
+enable_network_policy  = false
 
 notebook_nodes = {
   "small" : {
@@ -37,7 +40,6 @@ notebook_nodes = {
     max : 20,
     machine_type : "n1-standard-64"
   },
-
 }
 
 # Setup a single node pool for dask workers.

terraform/gcp/projects/pilot-hubs.tfvars

@@ -1,15 +1,19 @@
 prefix     = "pilot-hubs"
 project_id = "two-eye-two-see"
 
-zone   = "us-central1-b"
-region = "us-central1"
-
-core_node_machine_type = "n1-highmem-4"
+zone             = "us-central1-b"
+region           = "us-central1"
+regional_cluster = false
 
-# Multi-tenant cluster, network policy is required to enforce separation between hubs
-enable_network_policy = true
+k8s_versions = {
+  min_master_version : "1.26.5-gke.2100",
+  core_nodes_version : "1.26.5-gke.2100",
+  notebook_nodes_version : "1.26.4-gke.1400",
+  dask_nodes_version : "1.26.5-gke.1400",
+}
 
-regional_cluster = false
+core_node_machine_type = "n2-highmem-4"
+enable_network_policy  = true
 
 enable_filestore      = true
 filestore_capacity_gb = 5120

terraform/gcp/projects/qcl.tfvars

@@ -1,13 +1,19 @@
 prefix     = "qcl"
 project_id = "qcl-hub"
 
-zone   = "europe-west1-d"
-region = "europe-west1"
+zone             = "europe-west1-d"
+region           = "europe-west1"
+regional_cluster = true
+
+k8s_versions = {
+  min_master_version : "1.25.10-gke.2700",
+  core_nodes_version : "1.24.9-gke.3200",
+  notebook_nodes_version : "1.24.9-gke.3200",
+}
 
 core_node_machine_type = "n2-highmem-2"
 enable_network_policy  = true
 
-# Setup a filestore for in-cluster NFS
 enable_filestore      = true
 filestore_capacity_gb = 2048