Merge pull request #5179 from GeorgianaElena/gpu-prefix

AWS: GPU nodegroups and docs
2i2c-org · Nov 22, 2024 · 5079562 · 5079562
2 parents 9797232 + 7865fce
commit 5079562
Show file tree

Hide file tree

Showing 7 changed files with 23 additions and 18 deletions.
diff --git a/docs/howto/features/gpu.md b/docs/howto/features/gpu.md
@@ -112,8 +112,12 @@ AWS, and we can configure a node group there to provide us GPUs.
    ```
     {
         instanceType: "g4dn.xlarge",
+        namePrefix: "gpu-{{hub-name}}",
+        minSize: 0,
+        labels+: { "2i2c/hub-name": "{{hub-name}}" },
         tags+: {
-            "k8s.io/cluster-autoscaler/node-template/resources/nvidia.com/gpu": "1"
+            "k8s.io/cluster-autoscaler/node-template/resources/nvidia.com/gpu": "1",
+            "2i2c:hub-name": "{{hub-name}}",
         },
         taints+: {
             "nvidia.com/gpu": "present:NoSchedule"
@@ -126,7 +130,8 @@ AWS, and we can configure a node group there to provide us GPUs.
 
    `g4dn.xlarge` gives us 1 Nvidia T4 GPU and ~4 CPUs. The `tags` definition
    is necessary to let the autoscaler know that this nodegroup has
-   1 GPU per node. The `taints` definition is required to prevent scheduling of
+   1 GPU per node and also for the cost attribution system to differentiate
+   between hubs. The `taints` definition is required to prevent scheduling of
    non-GPU pods onto the GPU nodes. If you're using a different machine type with
    more GPUs, adjust this definition accordingly.
 

diff --git a/eksctl/2i2c-aws-us.jsonnet b/eksctl/2i2c-aws-us.jsonnet
@@ -84,7 +84,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "nb-showcase",
+        namePrefix: "gpu-showcase",
         minSize: 0,
         labels+: { "2i2c/hub-name": "showcase" },
         tags+: {
@@ -119,7 +119,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "nb-ncar-cisl",
+        namePrefix: "gpu-ncar-cisl",
         minSize: 0,
         labels+: { "2i2c/hub-name": "ncar-cisl" },
         tags+: {

diff --git a/eksctl/gridsst.jsonnet b/eksctl/gridsst.jsonnet
@@ -132,7 +132,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.xlarge",
         minSize: 0,
-        namePrefix: "nb-staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         tags+: {
             "2i2c:hub-name": "staging",
@@ -148,7 +148,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.xlarge",
         minSize: 0,
-        namePrefix: "nb-prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         tags+: {
             "2i2c:hub-name": "prod",

diff --git a/eksctl/jupyter-meets-the-earth.jsonnet b/eksctl/jupyter-meets-the-earth.jsonnet
@@ -112,7 +112,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.xlarge",
         minSize: 0,
-        namePrefix: "nb-staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         tags+: {
             "2i2c:hub-name": "staging",
@@ -125,7 +125,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.xlarge",
         minSize: 0,
-        namePrefix: "nb-prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         tags+: {
             "2i2c:hub-name": "prod",
@@ -138,7 +138,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.4xlarge",
         minSize: 0,
-        namePrefix: "nb-staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         tags+: {
             "2i2c:hub-name": "staging",
@@ -151,7 +151,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.4xlarge",
         minSize: 0,
-        namePrefix: "nb-prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         tags+: {
             "2i2c:hub-name": "prod",
@@ -164,7 +164,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.16xlarge",
         minSize: 0,
-        namePrefix: "nb-staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         taints+: {
           "nvidia.com/gpu": "NoSchedule"
@@ -177,7 +177,7 @@ local notebookNodes = [
     {
         instanceType: "g4dn.16xlarge",
         minSize: 0,
-        namePrefix: "nb-prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         taints+: {
           "nvidia.com/gpu": "NoSchedule"

diff --git a/eksctl/kitware.jsonnet b/eksctl/kitware.jsonnet
@@ -63,7 +63,7 @@ local notebookNodes = [
     },
     {
       instanceType: "g4dn.xlarge",
-      namePrefix: "nb-staging",
+      namePrefix: "gpu-staging",
       labels+: { "2i2c/hub-name": "staging" },
       tags+: {
         "2i2c:hub-name": "staging",
@@ -78,7 +78,7 @@ local notebookNodes = [
     },
     {
       instanceType: "g4dn.xlarge",
-      namePrefix: "nb-prod",
+      namePrefix: "gpu-prod",
       labels+: { "2i2c/hub-name": "prod" },
       tags+: {
         "2i2c:hub-name": "prod",

diff --git a/eksctl/nasa-cryo.jsonnet b/eksctl/nasa-cryo.jsonnet
@@ -63,7 +63,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         tags+: {
             "2i2c:hub-name": "staging",
@@ -78,7 +78,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         tags+: {
             "2i2c:hub-name": "prod",

diff --git a/eksctl/smithsonian.jsonnet b/eksctl/smithsonian.jsonnet
@@ -63,7 +63,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "nb-staging",
+        namePrefix: "gpu-staging",
         labels+: { "2i2c/hub-name": "staging" },
         tags+: {
             "2i2c:hub-name": "staging",
@@ -78,7 +78,7 @@ local notebookNodes = [
     },
     {
         instanceType: "g4dn.xlarge",
-        namePrefix: "nb-prod",
+        namePrefix: "gpu-prod",
         labels+: { "2i2c/hub-name": "prod" },
         tags+: {
             "2i2c:hub-name": "prod",