Training gateway plugin (#718)

* Introduce gateway plugin for model training

* Update protobuf definitions for WorkloadResponse

* Update proto files to reflect workload log count

* Update model_training proto file as well as api, plugin and system files

* Update modifications to api.go, plugin.go and system.go files

* Update plugin code

* Update grpc proto file for model training

* Update plugin for model training to address bug

* Update plugin

* Update API for training a model

* Update proto files for training controller gateway

* Update proto files for training controller gateway

* Add endpoint for checking GPU

* Update model training api to search for GPU

* Add opensearch aggregation routine as part of training controller gateway plugin

* Update model training plugin to include service for aggregating data from Opensearch

* Rename directory from modelTraining camel case to modeltraining

* Update model training directory names

* Fix up modeltraining plugin imports

* Update main.go file to reflect latest naming style

* Undo changes to opensearch image in repository

* Update plugin to fetch Opensearch endpoint and credentials

* Update code to make sure Opensearch credentials are properly being fetched

* Remove logging debug statements

* Update code to address PR comments

* Remove unused structs

* Update code to follow gofmt

* Update proto to rename GPUInfo list to items

* Update proto files with PR comments

* Update aggregation service to use context from ModelTrainingPlugin

* Update proto file definitions to be more distinct for each endpoint

* Add error handling for api endpoints and update proto definitions

* Update plugin file to check if os-workload-aggregation already exists in Jetstream

* Update proto definitions to be more concise

* Update struct field formatting to make it consistent

* Update composite aggregation size to be from 4 records per scroll to 1000 records per scroll

pkg/resources/gateway/rbac.go

@@ -118,7 +118,7 @@ func (r *Reconciler) rbac() ([]resources.Resource, error) {
 			},
 			{
 				APIGroups: []string{""},
-				Resources: []string{"endpoints"},
+				Resources: []string{"endpoints", "nodes"},
 				Verbs: []string{
 					"get",
 					"list",

plugins/modeltraining/main.go

@@ -0,0 +1,21 @@
+package main
+
+import (
+	"context"
+	"time"
+
+	"github.com/gin-gonic/gin"
+	"github.com/rancher/opni/pkg/plugins"
+	"github.com/rancher/opni/pkg/tracing"
+	"github.com/rancher/opni/pkg/util/waitctx"
+	modeltraining "github.com/rancher/opni/plugins/modeltraining/pkg/modeltraining"
+)
+
+func main() {
+	tracing.Configure("plugin_modeltraining")
+	gin.SetMode(gin.ReleaseMode)
+	ctx, ca := context.WithCancel(waitctx.Background())
+	plugins.Serve(modeltraining.Scheme(ctx))
+	ca()
+	waitctx.Wait(ctx, 5*time.Second)
+}