Select TPU by topology (#525)

* fix yaml formatting for workloads with TPU and NAP

* refactor tpu system characteristics

* fix tests

internal representation of TPU machines has changed, so now grep that used the old format fails

---------

Co-authored-by: pawloch00 <ppawl@google.com>

Author: gcie

.github/workflows/build_tests.yaml

@@ -20,10 +20,10 @@ on:
       tpu-type:
         description: 'TPU Type'
         required: true
-        default: 'v4-8'
+        default: 'v4-2x2x1'
         type: choice
         options:
-          - v4-8
+          - v4-2x2x1
   push:
     branches: ["main","develop"]
   pull_request: # By default this runs for types assigned, opened and synchronize.
@@ -70,7 +70,7 @@ jobs:
     - name: set tpu-type
       id: set-tpu-type
       run: |
-        echo tpu-type=v4-8 >> $GITHUB_OUTPUT
+        echo tpu-type=v4-2x2x1 >> $GITHUB_OUTPUT
     - name: set location
       id: set-location
       run: |    

.github/workflows/reusable_cluster_create.yaml

@@ -23,7 +23,7 @@ on:
       tpu-type:
         description: 'TPU Type'
         required: false
-        default: 'v4-8'
+        default: 'v4-2x2x1'
         type: string
       zone:
         type: string

.github/workflows/reusable_cluster_private.yaml

@@ -23,7 +23,7 @@ on:
       tpu-type:
         description: 'TPU Type'
         required: false
-        default: 'v4-8'
+        default: 'v4-2x2x1'
         type: string
       zone:
         type: string

.github/workflows/reusable_storage_create.yaml

@@ -21,7 +21,7 @@ on:
       tpu-type:
         description: 'TPU Type'
         required: false
-        default: 'v4-8'
+        default: 'v4-2x2x1'
         type: string
       zone:
         type: string

.github/workflows/reusable_storage_tests.yaml

@@ -23,7 +23,7 @@ on:
       tpu-type:
         description: 'TPU Type'
         required: false
-        default: 'v4-8'
+        default: 'v4-2x2x1'
         type: string
       zone:
         type: string

src/xpk/core/capacity.py

@@ -174,7 +174,7 @@ def get_capacity_node_selectors_from_capacity_type(
     case CapacityType.ON_DEMAND.name:
       node_selector = ''
     case CapacityType.SPOT.name:
-      node_selector = 'cloud.google.com/gke-spot="true"'
+      node_selector = 'cloud.google.com/gke-spot: "true"'
     case CapacityType.RESERVATION.name:
       node_selector = f'cloud.google.com/reservation-name: {args.reservation}'
     case _: