added accelerator detection

backend/src/routes/api/accelerators/acceleratorUtils.ts

@@ -0,0 +1,46 @@
+import { AcceleratorInfo, KubeFastifyInstance } from "../../../types"
+
+const RESOURCE_TYPES = ["cpu", "memory", "pods", "ephemeral-storage", "hugepages-1Gi", "hugepages-2Mi", "attachable-volumes-aws-ebs"]
+
+const getIdentifiersFromResources = (resources: {[key: string]: string} = {}) => {
+	return Object.entries(resources)
+	.filter(([key,]) => !RESOURCE_TYPES.includes(key))
+	.reduce<{[key: string]: number}>((identifiers, [key, value]) => {
+		identifiers[key] = isNaN(parseInt(value)) ? 0 : parseInt(value)
+		return identifiers
+	}, {})
+}
+
+export const getAcceleratorNumbers = async (fastify: KubeFastifyInstance): Promise<AcceleratorInfo> => (
+	fastify.kube.coreV1Api.listNode()
+	.then((res) => res.body.items.reduce<AcceleratorInfo>((info, node) => {
+		// reduce resources down to just the accelerators and their counts
+		const allocatable = getIdentifiersFromResources(node.status.allocatable)
+		const capacity = getIdentifiersFromResources(node.status.capacity)
+
+		// update the max count for each accelerator
+		Object.entries(allocatable).forEach(([key, value]) => (
+			info.available[key] = Math.max((info.available[key] || 0), value)
+		))
+
+		// update the total count for each accelerator
+		Object.entries(capacity).forEach(([key, value]) => (
+			info.total[key] = (info.total[key] || 0) + value
+		))
+
+
+		// update the allocated count for each accelerator
+		Object.entries(capacity).forEach(([key, value]) => (
+			info.allocated[key] = (info.allocated[key] || 0) + value - (allocatable[key] || 0)
+		))
+
+		// if any accelerators are available, the cluster is configured
+		const configured = info.configured || Object.values(info.available).some((value) => value > 0)
+
+		return {total: info.total, available: info.available, allocated: info.allocated, configured}
+	}, {configured: false, available: {}, total: {}, allocated: {}}))
+	.catch((e) => {
+		fastify.log.error(`Exception when listing cluster nodes: ${e}`);
+		return  {configured: false, available: {}, total: {}, allocated: {}}
+	})
+)

backend/src/routes/api/accelerators/index.ts

@@ -0,0 +1,11 @@
+import { KubeFastifyInstance, OauthFastifyRequest } from '../../../types';
+import { getAcceleratorNumbers } from './acceleratorUtils';
+import { logRequestDetails } from '../../../utils/fileUtils';
+
+export default async (fastify: KubeFastifyInstance): Promise<void> => {
+  fastify.get('/', async (request: OauthFastifyRequest) => {
+    logRequestDetails(fastify, request);
+
+    return getAcceleratorNumbers(fastify);
+  });
+};

backend/src/routes/api/gpu/gpuUtils.ts

@@ -16,6 +16,9 @@ const storage: { lastFetch: number; lastValue: GPUInfo } = {
   lastFetch: 0,
 };
 
+/**
+ * @deprecated - use getAcceleratorNumbers instead
+ */
 export const getGPUNumber = async (fastify: KubeFastifyInstance): Promise<GPUInfo> => {
   if (storage.lastFetch >= Date.now() - 30_000) {
     fastify.log.info(`Returning cached gpu value (${JSON.stringify(storage)})`);

backend/src/routes/api/gpu/index.ts

@@ -2,6 +2,9 @@ import { KubeFastifyInstance, OauthFastifyRequest } from '../../../types';
 import { getGPUNumber } from './gpuUtils';
 import { logRequestDetails } from '../../../utils/fileUtils';
 
+/**
+ * @deprecated - use accelerators instead
+ */
 export default async (fastify: KubeFastifyInstance): Promise<void> => {
   fastify.get('/', async (request: OauthFastifyRequest) => {
     logRequestDetails(fastify, request);

backend/src/types.ts

@@ -748,6 +748,14 @@ export type GPUInfo = {
   available: number;
   autoscalers: gpuScale[];
 };
+
+export type AcceleratorInfo = {
+  configured: boolean;
+  available: {[key: string]: number};
+  total: {[key: string]: number};
+  allocated: {[key: string]: number};
+}
+
 export type EnvironmentVariable = EitherNotBoth<
   { value: string | number },
   { valueFrom: Record<string, unknown> }