feat: initial release of headlamp-intel-gpu-plugin v0.1.0

Adds a Headlamp plugin for Intel GPU device plugin visibility:

- Dedicated sidebar section: Overview, Device Plugins, GPU Nodes, GPU Pods
- Native Node detail page injection: GPU capacity, allocatable, utilization, active pods
- Native Pod detail page injection: per-container GPU resource requests/limits
- Native Nodes table: GPU Type and GPU Devices columns
- App bar health badge (hidden when plugin not installed)
- GpuDevicePlugin CRD monitoring (deviceplugin.intel.com/v1) with graceful
  degradation when CRD is not present
- Supports discrete (i915), Xe, and integrated GPU nodes via node labels
- 48 unit tests, TypeScript clean, 28 kB production bundle

Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>
This commit is contained in:
2026-02-18 17:58:49 -05:00
commit 41bf2aead4
18 changed files with 21053 additions and 0 deletions
+46
View File
@@ -0,0 +1,46 @@
/**
* AppBarGpuBadge — compact Intel GPU health indicator in the Headlamp app bar.
*
* Shows a status chip in the top navigation bar summarising GPU plugin health.
* Hides itself when no Intel GPU plugin is detected.
*/
import { StatusLabel } from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
export default function AppBarGpuBadge() {
const { pluginInstalled, gpuNodes, devicePlugins, loading } = useIntelGpuContext();
// Hide when loading or no plugin present
if (loading || !pluginInstalled) return null;
const hasUnhealthyPlugin = devicePlugins.some(p => {
const desired = p.status?.desiredNumberScheduled ?? 0;
const ready = p.status?.numberReady ?? 0;
const unavailable = p.status?.numberUnavailable ?? 0;
return (desired > 0 && ready < desired) || unavailable > 0;
});
const status = hasUnhealthyPlugin ? 'warning' : 'success';
const nodeCount = gpuNodes.length;
return (
<div
style={{
display: 'flex',
alignItems: 'center',
gap: '4px',
padding: '0 8px',
cursor: 'default',
}}
title={`Intel GPU: ${nodeCount} node${nodeCount !== 1 ? 's' : ''}`}
>
<StatusLabel status={status}>
<span style={{ fontSize: '11px', fontWeight: 600 }}>
Intel GPU{nodeCount > 0 ? ` · ${nodeCount}N` : ''}
</span>
</StatusLabel>
</div>
);
}
+212
View File
@@ -0,0 +1,212 @@
/**
* DevicePluginsPage — lists all GpuDevicePlugin CRD instances.
*
* Shows configuration details for each Intel GPU device plugin deployment,
* including spec and status information.
*/
import {
Loader,
NameValueTable,
SectionBox,
SectionHeader,
SimpleTable,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
import { formatAge, isPodReady, pluginStatusText, pluginStatusToStatus } from '../api/k8s';
export default function DevicePluginsPage() {
const { devicePlugins, pluginPods, crdAvailable, loading, error, refresh } =
useIntelGpuContext();
if (loading) {
return <Loader title="Loading device plugin data..." />;
}
return (
<>
<div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: '20px' }}>
<SectionHeader title="Intel GPU — Device Plugins" />
<button
onClick={refresh}
aria-label="Refresh device plugin data"
style={{
padding: '6px 16px',
backgroundColor: 'transparent',
color: 'var(--mui-palette-primary-main, #0071c5)',
border: '1px solid var(--mui-palette-primary-main, #0071c5)',
borderRadius: '4px',
cursor: 'pointer',
fontSize: '13px',
fontWeight: 500,
}}
>
Refresh
</button>
</div>
{error && (
<SectionBox title="Error">
<NameValueTable
rows={[{ name: 'Status', value: <StatusLabel status="error">{error}</StatusLabel> }]}
/>
</SectionBox>
)}
{!crdAvailable && (
<SectionBox title="CRD Not Available">
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status="warning">
GpuDevicePlugin CRD (deviceplugin.intel.com/v1) is not installed
</StatusLabel>
),
},
{
name: 'Note',
value:
'Install the Intel Device Plugins Operator to manage GpuDevicePlugin resources. ' +
'Plugin daemon pods are shown below if detected.',
},
]}
/>
</SectionBox>
)}
{/* GpuDevicePlugin CRD instances */}
{crdAvailable && devicePlugins.length === 0 && (
<SectionBox title="No Device Plugins">
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status="warning">
No GpuDevicePlugin resources found on this cluster
</StatusLabel>
),
},
{
name: 'Create',
value:
'kubectl apply -f gpudeviceplugin.yaml (see Intel documentation for configuration)',
},
]}
/>
</SectionBox>
)}
{devicePlugins.map(plugin => (
<SectionBox key={plugin.metadata.uid ?? plugin.metadata.name} title={`GpuDevicePlugin: ${plugin.metadata.name}`}>
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status={pluginStatusToStatus(plugin)}>
{pluginStatusText(plugin)}
</StatusLabel>
),
},
{
name: 'Image',
value: plugin.spec.image ?? '—',
},
{
name: 'Shared Devices/Node',
value: String(plugin.spec.sharedDevNum ?? 1),
},
{
name: 'Allocation Policy',
value: plugin.spec.preferredAllocationPolicy ?? 'default',
},
{
name: 'Monitoring',
value: plugin.spec.enableMonitoring ? (
<StatusLabel status="success">Enabled</StatusLabel>
) : (
<StatusLabel status="warning">Disabled</StatusLabel>
),
},
{
name: 'Resource Manager',
value: plugin.spec.resourceManager ? 'Enabled' : 'Disabled',
},
{
name: 'Desired Nodes',
value: String(plugin.status?.desiredNumberScheduled ?? '—'),
},
{
name: 'Ready Nodes',
value: String(plugin.status?.numberReady ?? '—'),
},
...(plugin.status?.numberUnavailable
? [{
name: 'Unavailable Nodes',
value: (
<StatusLabel status="error">
{plugin.status.numberUnavailable}
</StatusLabel>
),
}]
: []),
{
name: 'Node Selector',
value: plugin.spec.nodeSelector
? Object.entries(plugin.spec.nodeSelector)
.map(([k, v]) => `${k}=${v}`)
.join(', ')
: '—',
},
{
name: 'Age',
value: formatAge(plugin.metadata.creationTimestamp),
},
]}
/>
</SectionBox>
))}
{/* Plugin daemon pods */}
{pluginPods.length > 0 && (
<SectionBox title="Plugin Daemon Pods">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{ label: 'Namespace', getter: (p) => p.metadata.namespace ?? '—' },
{ label: 'Node', getter: (p) => p.spec?.nodeName ?? '—' },
{
label: 'Ready',
getter: (p) => (
<StatusLabel status={isPodReady(p) ? 'success' : 'warning'}>
{isPodReady(p) ? 'Ready' : p.status?.phase ?? 'Unknown'}
</StatusLabel>
),
},
{
label: 'Restarts',
getter: (p) => {
const restarts = p.status?.containerStatuses?.reduce(
(sum, c) => sum + c.restartCount, 0
) ?? 0;
return restarts > 0 ? (
<StatusLabel status="warning">{restarts}</StatusLabel>
) : (
String(restarts)
);
},
},
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={pluginPods}
/>
</SectionBox>
)}
</>
);
}
+144
View File
@@ -0,0 +1,144 @@
/**
* NodeDetailSection — injected into Headlamp's native Node detail page.
*
* Shows Intel GPU resources available on the node (capacity, allocatable),
* GPU type, and pods currently using GPU resources on this node.
* Returns null for non-GPU nodes.
*/
import {
NameValueTable,
SectionBox,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
import {
formatGpuResourceName,
formatGpuType,
getGpuResources,
getNodeGpuType,
INTEL_GPU_RESOURCE,
INTEL_GPU_RESOURCE_PREFIX,
INTEL_GPU_XE_RESOURCE,
isIntelGpuNode,
isNodeReady,
} from '../api/k8s';
interface NodeDetailSectionProps {
resource: {
kind?: string;
metadata?: { name?: string; labels?: Record<string, string> };
jsonData?: unknown;
// Headlamp KubeObject may expose status directly or via jsonData
status?: unknown;
};
}
export default function NodeDetailSection({ resource }: NodeDetailSectionProps) {
const { gpuPods, loading } = useIntelGpuContext();
// Extract the raw Kubernetes JSON — Headlamp KubeObject wraps it in jsonData
const rawNode =
resource.jsonData && typeof resource.jsonData === 'object'
? resource.jsonData
: resource;
// Only render for Node resources that have Intel GPU
if (!isIntelGpuNode(rawNode)) return null;
const node = rawNode as Parameters<typeof isIntelGpuNode>[0] & {
status?: {
capacity?: Record<string, string>;
allocatable?: Record<string, string>;
nodeInfo?: { kernelVersion?: string; osImage?: string };
};
metadata: { name: string; labels?: Record<string, string> };
};
const nodeName = (node as { metadata: { name: string } }).metadata.name;
const capacity = getGpuResources((node as any).status?.capacity);
const allocatable = getGpuResources((node as any).status?.allocatable);
const gpuType = getNodeGpuType(node as any);
// Find GPU pods scheduled on this node
const podsOnNode = loading
? []
: gpuPods.filter(p => p.spec?.nodeName === nodeName);
if (Object.keys(capacity).length === 0 && Object.keys(allocatable).length === 0) {
return null;
}
// GPU utilization: count GPU units used by running pods
let gpuInUse = 0;
let gpuAllocatable = 0;
for (const [key, val] of Object.entries(allocatable)) {
if (key === INTEL_GPU_RESOURCE || key === INTEL_GPU_XE_RESOURCE) {
gpuAllocatable += parseInt(val, 10) || 0;
}
}
for (const pod of podsOnNode.filter(p => p.status?.phase === 'Running')) {
const reqs = pod.spec?.containers?.flatMap(c =>
Object.entries(c.resources?.requests ?? {}).filter(([k]) =>
k === INTEL_GPU_RESOURCE || k === INTEL_GPU_XE_RESOURCE
)
) ?? [];
for (const [, val] of reqs) {
gpuInUse += parseInt(val, 10) || 0;
}
}
const utilizationPct =
gpuAllocatable > 0 ? Math.round((gpuInUse / gpuAllocatable) * 100) : 0;
const utilizationStatus: 'success' | 'warning' | 'error' =
utilizationPct >= 90 ? 'error' : utilizationPct >= 70 ? 'warning' : 'success';
return (
<SectionBox title="Intel GPU">
<NameValueTable
rows={[
{
name: 'GPU Type',
value: formatGpuType(gpuType),
},
// Capacity rows
...Object.entries(capacity).map(([key, val]) => ({
name: `${formatGpuResourceName(key)} (capacity)`,
value: val,
})),
// Allocatable rows
...Object.entries(allocatable).map(([key, val]) => ({
name: `${formatGpuResourceName(key)} (allocatable)`,
value: val,
})),
// Utilization
...(gpuAllocatable > 0
? [
{
name: 'GPU Utilization',
value: (
<StatusLabel status={utilizationStatus}>
{`${gpuInUse}/${gpuAllocatable} (${utilizationPct}%)`}
</StatusLabel>
),
},
]
: []),
// Workload pods
{
name: 'GPU Workload Pods',
value:
podsOnNode.length > 0
? podsOnNode.map(p => p.metadata.name).join(', ')
: loading
? 'Loading…'
: 'None',
},
]}
/>
</SectionBox>
);
}
+297
View File
@@ -0,0 +1,297 @@
/**
* NodesPage — lists all nodes with Intel GPU capabilities.
*
* Shows GPU type, device count, resource allocation, and pod assignments
* for each GPU-capable node in the cluster.
*/
import {
Loader,
NameValueTable,
SectionBox,
SectionHeader,
SimpleTable,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
import {
formatAge,
formatGpuResourceName,
formatGpuType,
getGpuResources,
getNodeGpuCount,
getNodeGpuType,
INTEL_GPU_RESOURCE,
INTEL_GPU_RESOURCE_PREFIX,
INTEL_GPU_XE_RESOURCE,
IntelGpuNode,
isNodeReady,
} from '../api/k8s';
// ---------------------------------------------------------------------------
// GPU allocation bar component
// ---------------------------------------------------------------------------
function GpuAllocationBar({
used,
allocatable,
}: {
used: number;
allocatable: number;
}) {
if (allocatable === 0) return <span></span>;
const pct = Math.min(100, Math.round((used / allocatable) * 100));
const color = pct >= 90 ? '#d32f2f' : pct >= 70 ? '#f57c00' : '#0071c5';
return (
<div style={{ display: 'flex', alignItems: 'center', gap: '8px' }}>
<div
style={{
width: '80px',
height: '8px',
backgroundColor: '#e0e0e0',
borderRadius: '4px',
overflow: 'hidden',
}}
>
<div
style={{
width: `${pct}%`,
height: '100%',
backgroundColor: color,
borderRadius: '4px',
}}
/>
</div>
<span style={{ fontSize: '12px' }}>{`${used}/${allocatable} (${pct}%)`}</span>
</div>
);
}
// ---------------------------------------------------------------------------
// Node detail card
// ---------------------------------------------------------------------------
function NodeDetailCard({
node,
podsByNode,
}: {
node: IntelGpuNode;
podsByNode: Map<string, string[]>;
}) {
const gpuType = getNodeGpuType(node);
const gpuCount = getNodeGpuCount(node);
const ready = isNodeReady(node);
const capacityResources = getGpuResources(node.status?.capacity);
const allocatableResources = getGpuResources(node.status?.allocatable);
const podsOnNode = podsByNode.get(node.metadata.name) ?? [];
return (
<SectionBox title={node.metadata.name}>
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status={ready ? 'success' : 'error'}>
{ready ? 'Ready' : 'Not Ready'}
</StatusLabel>
),
},
{
name: 'GPU Type',
value: formatGpuType(gpuType),
},
...(gpuCount > 0
? [{ name: 'GPU Devices (i915/xe)', value: String(gpuCount) }]
: []),
...Object.entries(capacityResources).map(([key, cap]) => {
const alloc = parseInt(allocatableResources[key] ?? '0', 10);
const total = parseInt(cap, 10);
return {
name: `${formatGpuResourceName(key)} (capacity)`,
value: String(total),
};
}),
...Object.entries(allocatableResources).map(([key, alloc]) => {
return {
name: `${formatGpuResourceName(key)} (allocatable)`,
value: alloc ?? '0',
};
}),
{
name: 'GPU Workload Pods',
value: podsOnNode.length > 0 ? podsOnNode.join(', ') : '—',
},
{
name: 'OS Image',
value: node.status?.nodeInfo?.osImage ?? '—',
},
{
name: 'Kernel',
value: node.status?.nodeInfo?.kernelVersion ?? '—',
},
{
name: 'Kubelet',
value: node.status?.nodeInfo?.kubeletVersion ?? '—',
},
{
name: 'Age',
value: formatAge(node.metadata.creationTimestamp),
},
]}
/>
</SectionBox>
);
}
// ---------------------------------------------------------------------------
// Main component
// ---------------------------------------------------------------------------
export default function NodesPage() {
const { gpuNodes, gpuPods, loading, error, refresh } = useIntelGpuContext();
if (loading) {
return <Loader title="Loading GPU node data..." />;
}
// Build map: nodeName → list of GPU pod names
const podsByNode = new Map<string, string[]>();
for (const pod of gpuPods) {
if (!pod.spec?.nodeName) continue;
const existing = podsByNode.get(pod.spec.nodeName) ?? [];
existing.push(pod.metadata.name);
podsByNode.set(pod.spec.nodeName, existing);
}
// Build table data for summary
const tableData = gpuNodes.map(node => {
const gpuType = getNodeGpuType(node);
const gpuCount = getNodeGpuCount(node);
const ready = isNodeReady(node);
const capacity = node.status?.capacity ?? {};
const allocatable = node.status?.allocatable ?? {};
let totalCapacity = 0;
let totalAllocatable = 0;
for (const key of Object.keys(capacity)) {
if (key === INTEL_GPU_RESOURCE || key === INTEL_GPU_XE_RESOURCE) {
totalCapacity += parseInt(capacity[key] ?? '0', 10);
totalAllocatable += parseInt(allocatable[key] ?? '0', 10);
}
}
const podsOnNode = podsByNode.get(node.metadata.name) ?? [];
return {
node,
gpuType,
gpuCount,
ready,
totalCapacity,
totalAllocatable,
podsOnNode,
};
});
return (
<>
<div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: '20px' }}>
<SectionHeader title="Intel GPU — Nodes" />
<button
onClick={refresh}
aria-label="Refresh node data"
style={{
padding: '6px 16px',
backgroundColor: 'transparent',
color: 'var(--mui-palette-primary-main, #0071c5)',
border: '1px solid var(--mui-palette-primary-main, #0071c5)',
borderRadius: '4px',
cursor: 'pointer',
fontSize: '13px',
fontWeight: 500,
}}
>
Refresh
</button>
</div>
{error && (
<SectionBox title="Error">
<NameValueTable
rows={[{ name: 'Status', value: <StatusLabel status="error">{error}</StatusLabel> }]}
/>
</SectionBox>
)}
{gpuNodes.length === 0 && (
<SectionBox title="No GPU Nodes Found">
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status="warning">
No nodes with Intel GPU resources or labels were found
</StatusLabel>
),
},
{
name: 'Note',
value:
'Nodes appear here when they have gpu.intel.com/* resources or Intel GPU node labels. ' +
'Ensure the Intel GPU device plugin and Node Feature Discovery are installed.',
},
]}
/>
</SectionBox>
)}
{/* Summary table */}
{gpuNodes.length > 0 && (
<SectionBox title="GPU Node Summary">
<SimpleTable
columns={[
{ label: 'Node', getter: (d) => d.node.metadata.name },
{
label: 'Ready',
getter: (d) => (
<StatusLabel status={d.ready ? 'success' : 'error'}>
{d.ready ? 'Ready' : 'Not Ready'}
</StatusLabel>
),
},
{ label: 'GPU Type', getter: (d) => formatGpuType(d.gpuType) },
{ label: 'GPU Devices', getter: (d) => String(d.gpuCount || '—') },
{
label: 'Allocation',
getter: (d) => (
<GpuAllocationBar
used={d.podsOnNode.length}
allocatable={d.totalAllocatable || d.gpuCount}
/>
),
},
{ label: 'GPU Pods', getter: (d) => String(d.podsOnNode.length) },
{ label: 'Age', getter: (d) => formatAge(d.node.metadata.creationTimestamp) },
]}
data={tableData}
/>
</SectionBox>
)}
{/* Per-node detail cards */}
{gpuNodes.map(node => (
<NodeDetailCard
key={node.metadata.uid ?? node.metadata.name}
node={node}
podsByNode={podsByNode}
/>
))}
</>
);
}
+379
View File
@@ -0,0 +1,379 @@
/**
* OverviewPage — main dashboard for the Intel GPU plugin.
*
* Shows: plugin health, GPU node summary, resource allocation overview,
* and pods requesting GPU resources.
*/
import {
Loader,
NameValueTable,
PercentageBar,
SectionBox,
SectionHeader,
SimpleTable,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
import {
formatAge,
formatGpuType,
getNodeGpuCount,
getNodeGpuType,
getPodGpuRequests,
INTEL_GPU_RESOURCE,
INTEL_GPU_RESOURCE_PREFIX,
INTEL_GPU_XE_RESOURCE,
isNodeReady,
isPodReady,
pluginStatusText,
pluginStatusToStatus,
} from '../api/k8s';
// ---------------------------------------------------------------------------
// GPU type distribution chart
// ---------------------------------------------------------------------------
function gpuTypeChartData(
discreteCount: number,
integratedCount: number,
unknownCount: number
): Array<{ name: string; value: number; fill: string }> {
const data = [];
if (discreteCount > 0) data.push({ name: 'Discrete', value: discreteCount, fill: '#0071c5' });
if (integratedCount > 0) data.push({ name: 'Integrated', value: integratedCount, fill: '#60a4dc' });
if (unknownCount > 0) data.push({ name: 'Unknown', value: unknownCount, fill: '#9e9e9e' });
return data;
}
// ---------------------------------------------------------------------------
// Main component
// ---------------------------------------------------------------------------
export default function OverviewPage() {
const {
devicePlugins,
pluginInstalled,
gpuNodes,
gpuPods,
pluginPods,
crdAvailable,
loading,
error,
refresh,
} = useIntelGpuContext();
if (loading) {
return <Loader title="Loading Intel GPU data..." />;
}
// Node type breakdown
let discreteCount = 0;
let integratedCount = 0;
let unknownCount = 0;
let totalGpuCount = 0;
let readyNodeCount = 0;
for (const node of gpuNodes) {
const type = getNodeGpuType(node);
if (type === 'discrete') discreteCount++;
else if (type === 'integrated') integratedCount++;
else unknownCount++;
totalGpuCount += getNodeGpuCount(node);
if (isNodeReady(node)) readyNodeCount++;
}
// GPU allocation summary: sum capacity vs allocatable across all GPU nodes
let totalCapacityGpus = 0;
let totalAllocatableGpus = 0;
let totalAllocatedGpus = 0;
for (const node of gpuNodes) {
const capacity = node.status?.capacity ?? {};
const allocatable = node.status?.allocatable ?? {};
for (const key of Object.keys(capacity)) {
if (key === INTEL_GPU_RESOURCE || key === INTEL_GPU_XE_RESOURCE) {
totalCapacityGpus += parseInt(capacity[key] ?? '0', 10);
totalAllocatableGpus += parseInt(allocatable[key] ?? '0', 10);
}
}
}
// Count GPUs in use from pods
for (const pod of gpuPods) {
if (pod.status?.phase !== 'Running') continue;
const requests = getPodGpuRequests(pod);
for (const [key, value] of Object.entries(requests)) {
if (key === INTEL_GPU_RESOURCE || key === INTEL_GPU_XE_RESOURCE) {
totalAllocatedGpus += parseInt(value, 10) || 0;
}
}
}
const gpuUtilizationPct =
totalCapacityGpus > 0
? Math.round((totalAllocatedGpus / totalCapacityGpus) * 100)
: 0;
const chartData = gpuTypeChartData(discreteCount, integratedCount, unknownCount);
const totalGpuNodes = gpuNodes.length;
// Pod phase breakdown
const podPhaseCounts = { Running: 0, Pending: 0, Succeeded: 0, Failed: 0, Other: 0 };
for (const pod of gpuPods) {
const phase = pod.status?.phase ?? 'Other';
if (phase in podPhaseCounts) {
podPhaseCounts[phase as keyof typeof podPhaseCounts]++;
} else {
podPhaseCounts.Other++;
}
}
return (
<>
<div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: '20px' }}>
<SectionHeader title="Intel GPU — Overview" />
<button
onClick={refresh}
aria-label="Refresh Intel GPU data"
style={{
padding: '6px 16px',
backgroundColor: 'transparent',
color: 'var(--mui-palette-primary-main, #0071c5)',
border: '1px solid var(--mui-palette-primary-main, #0071c5)',
borderRadius: '4px',
cursor: 'pointer',
fontSize: '13px',
fontWeight: 500,
}}
>
Refresh
</button>
</div>
{/* Error state */}
{error && (
<SectionBox title="Error">
<NameValueTable
rows={[{ name: 'Status', value: <StatusLabel status="error">{error}</StatusLabel> }]}
/>
</SectionBox>
)}
{/* Plugin not detected */}
{!pluginInstalled && !loading && (
<SectionBox title="Plugin Not Detected">
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status="warning">
Intel GPU device plugin not found on this cluster
</StatusLabel>
),
},
{
name: 'Install (Helm)',
value:
'helm repo add intel https://intel.github.io/helm-charts && ' +
'helm install intel-device-plugins-operator intel/intel-device-plugins-operator',
},
{
name: 'Documentation',
value: 'https://intel.github.io/intel-device-plugins-for-kubernetes/',
},
]}
/>
</SectionBox>
)}
{/* CRD not available notice */}
{!crdAvailable && pluginInstalled && (
<SectionBox title="Notice">
<NameValueTable
rows={[
{
name: 'CRD Status',
value: (
<StatusLabel status="warning">
GpuDevicePlugin CRD not found limited visibility available
</StatusLabel>
),
},
{
name: 'Note',
value:
'Plugin pods detected via DaemonSet labels. Install the Intel Device Plugins Operator for full CRD-based management.',
},
]}
/>
</SectionBox>
)}
{/* Device Plugin status — only shown when CRDs exist */}
{crdAvailable && devicePlugins.length > 0 && (
<SectionBox title="Device Plugin Status">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{
label: 'Status',
getter: (p) => (
<StatusLabel status={pluginStatusToStatus(p)}>
{pluginStatusText(p)}
</StatusLabel>
),
},
{
label: 'Monitoring',
getter: (p) => p.spec.enableMonitoring ? (
<StatusLabel status="success">Enabled</StatusLabel>
) : (
<StatusLabel status="warning">Disabled</StatusLabel>
),
},
{ label: 'Shared/Node', getter: (p) => String(p.spec.sharedDevNum ?? 1) },
{ label: 'Policy', getter: (p) => p.spec.preferredAllocationPolicy ?? '—' },
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={devicePlugins}
/>
</SectionBox>
)}
{/* Plugin daemon pods (shown when no CRD, or always as supplemental) */}
{pluginPods.length > 0 && (
<SectionBox title="Plugin Daemon Pods">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{ label: 'Namespace', getter: (p) => p.metadata.namespace ?? '—' },
{ label: 'Node', getter: (p) => p.spec?.nodeName ?? '—' },
{
label: 'Status',
getter: (p) => (
<StatusLabel status={isPodReady(p) ? 'success' : 'warning'}>
{isPodReady(p) ? 'Ready' : p.status?.phase ?? 'Unknown'}
</StatusLabel>
),
},
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={pluginPods}
/>
</SectionBox>
)}
{/* GPU Node summary */}
<SectionBox title="GPU Nodes">
{totalGpuNodes > 0 && chartData.length > 0 && (
<div style={{ marginBottom: '16px' }}>
<div style={{ marginBottom: '8px', fontSize: '14px', color: 'var(--mui-palette-text-secondary)' }}>
GPU Type Distribution
</div>
<PercentageBar data={chartData} total={totalGpuNodes} />
</div>
)}
<NameValueTable
rows={[
{
name: 'Total GPU Nodes',
value: (
<StatusLabel status={totalGpuNodes > 0 ? 'success' : 'warning'}>
{totalGpuNodes}
</StatusLabel>
),
},
{ name: 'Ready Nodes', value: String(readyNodeCount) },
...(discreteCount > 0 ? [{ name: 'Discrete GPU Nodes', value: String(discreteCount) }] : []),
...(integratedCount > 0 ? [{ name: 'Integrated GPU Nodes', value: String(integratedCount) }] : []),
...(totalGpuCount > 0 ? [{ name: 'Total GPU Devices', value: String(totalGpuCount) }] : []),
]}
/>
</SectionBox>
{/* GPU allocation summary */}
{totalCapacityGpus > 0 && (
<SectionBox title="GPU Allocation">
<div style={{ marginBottom: '16px' }}>
<div style={{ marginBottom: '8px', fontSize: '14px', color: 'var(--mui-palette-text-secondary)' }}>
GPU Utilization ({gpuUtilizationPct}%)
</div>
<PercentageBar
data={[
{ name: 'In Use', value: totalAllocatedGpus, fill: '#0071c5' },
{ name: 'Available', value: totalAllocatableGpus - totalAllocatedGpus, fill: '#e0e0e0' },
]}
total={totalAllocatableGpus}
/>
</div>
<NameValueTable
rows={[
{ name: 'Total Capacity (GPU devices)', value: String(totalCapacityGpus) },
{ name: 'Allocatable', value: String(totalAllocatableGpus) },
{ name: 'In Use', value: String(totalAllocatedGpus) },
{
name: 'Free',
value: (
<StatusLabel
status={totalAllocatableGpus - totalAllocatedGpus > 0 ? 'success' : 'warning'}
>
{totalAllocatableGpus - totalAllocatedGpus}
</StatusLabel>
),
},
]}
/>
</SectionBox>
)}
{/* GPU workloads summary */}
<SectionBox title="GPU Workloads">
<NameValueTable
rows={[
{ name: 'Total GPU Pods', value: String(gpuPods.length) },
...(podPhaseCounts.Running > 0
? [{ name: 'Running', value: <StatusLabel status="success">{podPhaseCounts.Running}</StatusLabel> }]
: []),
...(podPhaseCounts.Pending > 0
? [{ name: 'Pending', value: <StatusLabel status="warning">{podPhaseCounts.Pending}</StatusLabel> }]
: []),
...(podPhaseCounts.Failed > 0
? [{ name: 'Failed', value: <StatusLabel status="error">{podPhaseCounts.Failed}</StatusLabel> }]
: []),
]}
/>
</SectionBox>
{/* Active GPU pods list (running only, trimmed to top 10) */}
{gpuPods.filter(p => p.status?.phase === 'Running').length > 0 && (
<SectionBox title="Active GPU Pods">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{ label: 'Namespace', getter: (p) => p.metadata.namespace ?? '—' },
{ label: 'Node', getter: (p) => p.spec?.nodeName ?? '—' },
{
label: 'GPU Request',
getter: (p) => {
const reqs = getPodGpuRequests(p);
const parts: string[] = [];
for (const [key, val] of Object.entries(reqs)) {
const shortKey = key.replace(INTEL_GPU_RESOURCE_PREFIX, '');
parts.push(`${shortKey}: ${val}`);
}
return parts.join(', ') || '—';
},
},
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={gpuPods.filter(p => p.status?.phase === 'Running').slice(0, 10)}
/>
</SectionBox>
)}
</>
);
}
+118
View File
@@ -0,0 +1,118 @@
/**
* PodDetailSection — injected into Headlamp's native Pod detail page.
*
* Shows Intel GPU resource requests and limits per container, plus
* a link to the node's GPU summary.
* Returns null for pods that don't request Intel GPU resources.
*/
import {
NameValueTable,
SectionBox,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { formatGpuResourceName, INTEL_GPU_RESOURCE_PREFIX, isGpuRequestingPod } from '../api/k8s';
interface PodDetailSectionProps {
resource: {
kind?: string;
metadata?: { name?: string; namespace?: string };
jsonData?: unknown;
};
}
export default function PodDetailSection({ resource }: PodDetailSectionProps) {
// Extract raw Kubernetes JSON
const rawPod =
resource.jsonData && typeof resource.jsonData === 'object'
? resource.jsonData
: resource;
// Only render for pods that request Intel GPU resources
if (!isGpuRequestingPod(rawPod)) return null;
const pod = rawPod as {
metadata: { name: string; namespace?: string };
spec?: {
nodeName?: string;
containers?: Array<{
name: string;
resources?: {
requests?: Record<string, string>;
limits?: Record<string, string>;
};
}>;
};
status?: { phase?: string };
};
const containers = pod.spec?.containers ?? [];
const gpuContainers = containers.filter(c => {
const all = { ...c.resources?.requests, ...c.resources?.limits };
return Object.keys(all).some(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX));
});
if (gpuContainers.length === 0) return null;
// Build rows: one per container per GPU resource
const rows: Array<{ name: string; value: React.ReactNode }> = [];
for (const c of gpuContainers) {
const requests = c.resources?.requests ?? {};
const limits = c.resources?.limits ?? {};
const allGpuKeys = new Set([
...Object.keys(requests).filter(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX)),
...Object.keys(limits).filter(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX)),
]);
for (const key of allGpuKeys) {
const req = requests[key];
const lim = limits[key];
const resourceName = formatGpuResourceName(key);
rows.push({
name: `${c.name}${resourceName} request`,
value: req ?? '—',
});
if (lim && lim !== req) {
rows.push({
name: `${c.name}${resourceName} limit`,
value: lim,
});
}
}
}
const phase = pod.status?.phase;
const phaseStatus: 'success' | 'warning' | 'error' =
phase === 'Running' || phase === 'Succeeded'
? 'success'
: phase === 'Pending'
? 'warning'
: 'error';
return (
<SectionBox title="Intel GPU Resources">
<NameValueTable
rows={[
{
name: 'Phase',
value: (
<StatusLabel status={phaseStatus}>{phase ?? 'Unknown'}</StatusLabel>
),
},
{
name: 'Scheduled Node',
value: pod.spec?.nodeName ?? '—',
},
{
name: 'GPU Containers',
value: String(gpuContainers.length),
},
...rows,
]}
/>
</SectionBox>
);
}
+246
View File
@@ -0,0 +1,246 @@
/**
* PodsPage — lists all pods requesting Intel GPU resources.
*
* Shows GPU resource requests/limits per container and pod-level status.
*/
import {
Loader,
NameValueTable,
SectionBox,
SectionHeader,
SimpleTable,
StatusLabel,
} from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import { useIntelGpuContext } from '../api/IntelGpuDataContext';
import {
formatAge,
formatGpuResourceName,
IntelGpuPod,
INTEL_GPU_RESOURCE_PREFIX,
isPodReady,
getPodGpuRequests,
getPodRestarts,
} from '../api/k8s';
// ---------------------------------------------------------------------------
// Phase → status mapping
// ---------------------------------------------------------------------------
function phaseToStatus(phase: string | undefined): 'success' | 'warning' | 'error' {
switch (phase) {
case 'Running': return 'success';
case 'Succeeded': return 'success';
case 'Pending': return 'warning';
case 'Failed': return 'error';
default: return 'warning';
}
}
// ---------------------------------------------------------------------------
// GPU container list for a pod
// ---------------------------------------------------------------------------
function GpuContainerList({ pod }: { pod: IntelGpuPod }) {
const containers = pod.spec?.containers ?? [];
const gpuContainers = containers.filter(c => {
const resources = { ...c.resources?.requests, ...c.resources?.limits };
return Object.keys(resources).some(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX));
});
if (gpuContainers.length === 0) return <span></span>;
return (
<>
{gpuContainers.map(c => {
const requests = c.resources?.requests ?? {};
const limits = c.resources?.limits ?? {};
const gpuKeys = new Set([
...Object.keys(requests).filter(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX)),
...Object.keys(limits).filter(k => k.startsWith(INTEL_GPU_RESOURCE_PREFIX)),
]);
const parts: string[] = [];
for (const key of gpuKeys) {
const shortKey = formatGpuResourceName(key);
const req = requests[key];
const lim = limits[key];
if (req && lim && req === lim) {
parts.push(`${shortKey}: ${req}`);
} else if (req || lim) {
parts.push(`${shortKey}: req=${req ?? '—'} lim=${lim ?? '—'}`);
}
}
return (
<div key={c.name} style={{ marginBottom: '2px', fontSize: '13px' }}>
<strong>{c.name}</strong>: {parts.join(', ')}
</div>
);
})}
</>
);
}
// ---------------------------------------------------------------------------
// Main component
// ---------------------------------------------------------------------------
export default function PodsPage() {
const { gpuPods, loading, error, refresh } = useIntelGpuContext();
if (loading) {
return <Loader title="Loading GPU pod data..." />;
}
// Group by phase
const running = gpuPods.filter(p => p.status?.phase === 'Running');
const pending = gpuPods.filter(p => p.status?.phase === 'Pending');
const failed = gpuPods.filter(p => p.status?.phase === 'Failed');
const other = gpuPods.filter(
p => !['Running', 'Pending', 'Failed'].includes(p.status?.phase ?? '')
);
return (
<>
<div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: '20px' }}>
<SectionHeader title="Intel GPU — Pods" />
<button
onClick={refresh}
aria-label="Refresh pod data"
style={{
padding: '6px 16px',
backgroundColor: 'transparent',
color: 'var(--mui-palette-primary-main, #0071c5)',
border: '1px solid var(--mui-palette-primary-main, #0071c5)',
borderRadius: '4px',
cursor: 'pointer',
fontSize: '13px',
fontWeight: 500,
}}
>
Refresh
</button>
</div>
{error && (
<SectionBox title="Error">
<NameValueTable
rows={[{ name: 'Status', value: <StatusLabel status="error">{error}</StatusLabel> }]}
/>
</SectionBox>
)}
{gpuPods.length === 0 && (
<SectionBox title="No GPU Pods Found">
<NameValueTable
rows={[
{
name: 'Status',
value: (
<StatusLabel status="warning">
No pods requesting Intel GPU resources were found
</StatusLabel>
),
},
{
name: 'Note',
value:
'Pods appear here when they request resources like gpu.intel.com/i915 or gpu.intel.com/xe.',
},
]}
/>
</SectionBox>
)}
{/* Summary */}
{gpuPods.length > 0 && (
<SectionBox title="Summary">
<NameValueTable
rows={[
{ name: 'Total GPU Pods', value: String(gpuPods.length) },
...(running.length > 0
? [{ name: 'Running', value: <StatusLabel status="success">{running.length}</StatusLabel> }]
: []),
...(pending.length > 0
? [{ name: 'Pending', value: <StatusLabel status="warning">{pending.length}</StatusLabel> }]
: []),
...(failed.length > 0
? [{ name: 'Failed', value: <StatusLabel status="error">{failed.length}</StatusLabel> }]
: []),
]}
/>
</SectionBox>
)}
{/* All pods table */}
{gpuPods.length > 0 && (
<SectionBox title="All GPU Pods">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{ label: 'Namespace', getter: (p) => p.metadata.namespace ?? '—' },
{ label: 'Node', getter: (p) => p.spec?.nodeName ?? '—' },
{
label: 'Phase',
getter: (p) => (
<StatusLabel status={phaseToStatus(p.status?.phase)}>
{p.status?.phase ?? 'Unknown'}
</StatusLabel>
),
},
{
label: 'GPU Resources',
getter: (p) => <GpuContainerList pod={p} />,
},
{
label: 'Restarts',
getter: (p) => {
const restarts = getPodRestarts(p);
return restarts > 0 ? (
<StatusLabel status="warning">{restarts}</StatusLabel>
) : (
String(restarts)
);
},
},
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={gpuPods}
/>
</SectionBox>
)}
{/* Pending pods attention box */}
{pending.length > 0 && (
<SectionBox title="Attention: Pending GPU Pods">
<SimpleTable
columns={[
{ label: 'Name', getter: (p) => p.metadata.name },
{ label: 'Namespace', getter: (p) => p.metadata.namespace ?? '—' },
{
label: 'GPU Resources',
getter: (p) => {
const reqs = getPodGpuRequests(p);
return Object.entries(reqs)
.map(([k, v]) => `${formatGpuResourceName(k)}: ${v}`)
.join(', ') || '—';
},
},
{
label: 'Waiting Reason',
getter: (p) => {
const reason = p.status?.containerStatuses?.[0]?.state?.waiting?.reason;
return reason ?? '—';
},
},
{ label: 'Age', getter: (p) => formatAge(p.metadata.creationTimestamp) },
]}
data={pending}
/>
</SectionBox>
)}
</>
);
}
@@ -0,0 +1,58 @@
/**
* NodeColumns — adds Intel GPU columns to the native Headlamp Nodes table.
*
* Injects two columns:
* - "GPU Type" — Discrete / Integrated / — for non-GPU nodes
* - "GPU Devices" — count of i915/xe devices available on the node
*
* The processor is registered via registerResourceTableColumnsProcessor
* in index.tsx, targeting the 'headlamp-nodes' table ID.
*/
import { StatusLabel } from '@kinvolk/headlamp-plugin/lib/CommonComponents';
import React from 'react';
import {
formatGpuType,
getNodeGpuCount,
getNodeGpuType,
isIntelGpuNode,
} from '../../api/k8s';
/** Build GPU columns to append to the native Nodes table. */
export function buildNodeGpuColumns() {
return [
{
label: 'GPU Type',
getter: (resource: unknown) => {
// resource is a Headlamp KubeObject — extract jsonData
const raw =
resource && typeof resource === 'object' && 'jsonData' in resource
? (resource as { jsonData: unknown }).jsonData
: resource;
if (!isIntelGpuNode(raw)) return '—';
const node = raw as Parameters<typeof getNodeGpuType>[0];
const type = getNodeGpuType(node);
return (
<StatusLabel status="success">
{formatGpuType(type)}
</StatusLabel>
);
},
},
{
label: 'GPU Devices',
getter: (resource: unknown) => {
const raw =
resource && typeof resource === 'object' && 'jsonData' in resource
? (resource as { jsonData: unknown }).jsonData
: resource;
if (!isIntelGpuNode(raw)) return '—';
const node = raw as Parameters<typeof getNodeGpuCount>[0];
const count = getNodeGpuCount(node);
return count > 0 ? String(count) : '—';
},
},
];
}