feat: extend cluster k8s_options with multi-vendor manifest support

- New k8s_pod_spec form sections (image credentials, node selector, gpu vendor overrides) under k8s_options, replacing the legacy flat k8s_volume_mounts list. UI keys aligned to the backend wire shape (snake_case k8s_options + camelCase inside). - System default container registry is pre-filled into the first image credential when creating a new cluster; empty username/password are coerced to null on submit to match the Optional[str] backend schema. - Register cluster flow supports multi-runtime selection gated by the cluster's gpuVendorOverrides: non-override vendors stay single-select with an inline hint; multi-add only opens once an override vendor is picked, and non-override cards become disabled in that state. - Manifest URL emits multiple ?runtime= params; check-env step combines per-vendor commands; downstream steps are disabled when no vendor is selected. - Pre-validate gpuVendorOverrides at save time (non-empty selector, no duplicates across vendors, no key clash with base nodeSelector) so the user sees the error before hitting the manifest endpoint. - Misc: dark-mode background of the k8s_pod_spec / volume mount titles no longer clashes with the drawer; cluster Steps no longer leaks the internal showModules/showForms props to the DOM.
2026-05-26 16:54:14 +08:00
parent 91c3d0b814
commit acb90531b2
21 changed files with 1023 additions and 112 deletions
@@ -1,3 +1,4 @@
+import { systemConfigAtom } from '@/atoms/system';
 import PluginExtraFields from '@/components/plugin-extra-fields';
 import { PageAction } from '@/config';
 import { PageActionType } from '@/config/types';
@@ -8,7 +9,9 @@ import {
  Textarea as SealTextArea
 } from '@gpustack/core-ui';
 import { useIntl } from '@umijs/max';
-import { Form } from 'antd';
+import { Form, message } from 'antd';
+import { useAtomValue } from 'jotai';
+import _ from 'lodash';
 import React, { forwardRef, useEffect, useImperativeHandle } from 'react';
 import { ProviderType, ProviderValueMap } from '../config';
 import {
@@ -32,6 +35,7 @@ const ClusterForm: React.FC<AddModalProps> = forwardRef(
    const intl = useIntl();
    const [activeKey, setActiveKey] = React.useState<string[]>([]);
    const advanceConfigRef = React.useRef<any>(null);
+    const systemConfig = useAtomValue(systemConfigAtom);

    const handleOnCollapseChange = async (keys: string | string[]) => {
      setActiveKey(Array.isArray(keys) ? keys : [keys]);
@@ -50,20 +54,107 @@ const ClusterForm: React.FC<AddModalProps> = forwardRef(
      }
    }, [activeKey, action]);

-    const handleOnFinish = (values: FormData) => {
-      const workerConfig = yaml2Json(advanceConfigRef.current?.getYamlValue());
+    // Mirror the backend's `_validate_multi_vendor_overrides` check so the
+    // user sees the problem at save time instead of as a 400 when they
+    // run the register command. Returns the i18n'd message of the first
+    // problem found, or null when clean.
+    const validateGpuVendorOverrides = (values: any): string | null => {
+      const opts = values?.k8s_options;
+      const overrides = opts?.gpuVendorOverrides;
+      if (!overrides) return null;
+      const entries = Object.entries(overrides) as [string, any][];
+      if (entries.length === 0) return null;

-      onFinish({
-        ...values,
-        worker_config: {
-          ...workerConfig
+      // 1. Every override entry must declare a non-empty nodeSelector,
+      //    otherwise it can't actually pin its DaemonSet to anything.
+      for (const [vendor, override] of entries) {
+        const sel = override?.nodeSelector;
+        if (!sel || Object.keys(sel).length === 0) {
+          return intl.formatMessage(
+            { id: 'clusters.gpuVendorOverrides.validate.emptySelector' },
+            { vendor }
+          );
        }
-      });
+      }
+
+      // 2. Two vendors with the same selector would fight for the same
+      //    nodes — backend rejects this at manifest render.
+      for (let i = 0; i < entries.length; i++) {
+        for (let j = i + 1; j < entries.length; j++) {
+          const [v1, o1] = entries[i];
+          const [v2, o2] = entries[j];
+          if (_.isEqual(o1?.nodeSelector, o2?.nodeSelector)) {
+            return intl.formatMessage(
+              { id: 'clusters.gpuVendorOverrides.validate.duplicate' },
+              { v1, v2 }
+            );
+          }
+        }
+      }
+
+      // 3. Base nodeSelector keys can't be reused in any override —
+      //    the CPU worker would require AND forbid the same key.
+      const baseKeys = Object.keys(opts?.nodeSelector || {});
+      if (baseKeys.length > 0) {
+        for (const [vendor, override] of entries) {
+          const overrideKeys = Object.keys(override?.nodeSelector || {});
+          const clash = overrideKeys.filter((k) => baseKeys.includes(k));
+          if (clash.length > 0) {
+            return intl.formatMessage(
+              { id: 'clusters.gpuVendorOverrides.validate.keyConflict' },
+              { vendor, keys: clash.join(', ') }
+            );
+          }
+        }
+      }
+
+      return null;
+    };
+
+    // Empty username/password aren't meaningful as credentials — the backend
+    // models them as Optional[str] and treats null as "no auth". Coerce
+    // before sending so a public registry placeholder stays unambiguous.
+    const normalizeOutgoing = (values: any): any => {
+      const creds = values?.k8s_options?.imageCredentials;
+      if (!Array.isArray(creds)) return values;
+      const fixed = creds.map((c: any) => ({
+        ...c,
+        username: c?.username ? c.username : null,
+        password: c?.password ? c.password : null
+      }));
+      return {
+        ...values,
+        k8s_options: { ...values.k8s_options, imageCredentials: fixed }
+      };
+    };
+
+    const handleOnFinish = (_values: FormData) => {
+      const workerConfig = yaml2Json(advanceConfigRef.current?.getYamlValue());
+      // antd's onFinish only delivers values for registered Form.Items.
+      // Spreading those on top of `getFieldsValue(true)` clobbers nested
+      // objects (e.g. `k8s_options` would lose `gpuVendorOverrides`, which is
+      // only set via setFieldValue), so we go straight to the full store.
+      const fullValues = form.getFieldsValue(true);
+
+      const overridesErr = validateGpuVendorOverrides(fullValues);
+      if (overridesErr) {
+        message.error(overridesErr);
+        return;
+      }
+
+      onFinish(
+        normalizeOutgoing({
+          ...fullValues,
+          worker_config: {
+            ...workerConfig
+          }
+        })
+      );
    };

    useEffect(() => {
      if (currentData) {
-        const volumeMounts = currentData?.k8s_volume_mounts || [];
+        const volumeMounts = currentData?.k8s_options?.volumeMounts || [];
        const realVolumeList = (volumeMounts || []).map(
          (item: any, index: number) => ({
            ...item,
@@ -72,27 +163,40 @@ const ClusterForm: React.FC<AddModalProps> = forwardRef(
        );
        form.setFieldsValue({
          ...currentData,
-          k8s_volume_mounts: realVolumeList
+          k8s_options: {
+            ...(currentData?.k8s_options || {}),
+            volumeMounts: realVolumeList
+          }
        });
      } else {
+        const defaultRegistry = systemConfig?.system_default_container_registry;
        form.setFieldsValue({
-          k8s_volume_mounts: [
-            {
-              name: 'gpustack-data-dir',
-              mountPath: '/var/lib/gpustack',
-              readOnly: false,
-              sourceType: 'hostPath',
-              volumeSource: {
-                hostPath: {
-                  path: '/var/lib/gpustack',
-                  type: 'DirectoryOrCreate'
+          k8s_options: {
+            volumeMounts: [
+              {
+                name: 'gpustack-data-dir',
+                mountPath: '/var/lib/gpustack',
+                readOnly: false,
+                sourceType: 'hostPath',
+                volumeSource: {
+                  hostPath: {
+                    path: '/var/lib/gpustack',
+                    type: 'DirectoryOrCreate'
+                  }
                }
              }
-            }
-          ]
+            ],
+            ...(defaultRegistry
+              ? {
+                  imageCredentials: [
+                    { registry: defaultRegistry, username: '', password: '' }
+                  ]
+                }
+              : {})
+          }
        });
      }
-    }, [currentData]);
+    }, [currentData, systemConfig?.system_default_container_registry]);

    useEffect(() => {
      if (currentData) {
@@ -130,17 +234,33 @@ const ClusterForm: React.FC<AddModalProps> = forwardRef(
        };
      },
      validateFields: async () => {
-        const values = await form.validateFields();
+        // Run validation first to display any field errors. Then read the
+        // FULL store via `getFieldsValue(true)` so values that were set via
+        // setFieldValue on non-registered paths (e.g. gpuVendorOverrides)
+        // are still included in what we hand to the API.
+        await form.validateFields();
+        const values = form.getFieldsValue(true);
+
+        // Mirror the backend invariants for gpuVendorOverrides so the user
+        // sees the same constraint before submit instead of at register time.
+        const overridesErr = validateGpuVendorOverrides(values);
+        if (overridesErr) {
+          message.error(overridesErr);
+          // Reject so the step-flow's Promise.allSettled marks this form
+          // as failed and the outer onNext skips the submit callback.
+          throw new Error(overridesErr);
+        }
+
        const workerConfig = yaml2Json(
          advanceConfigRef.current?.getYamlValue()
        );

-        return {
+        return normalizeOutgoing({
          ...values,
          worker_config: {
            ...workerConfig
          }
-        };
+        });
      }
    }));

@@ -207,6 +327,7 @@ const ClusterForm: React.FC<AddModalProps> = forwardRef(
                <AdvanceConfig
                  action={action}
                  provider={provider}
+                  currentData={currentData}
                  ref={advanceConfigRef}
                ></AdvanceConfig>
              )