Webportal support jobType in job protocol and dedicated parameters in inference job (#113)

* add job type support; add inference job support

* update

* deploy webportal-dind when webportal changed

* update

* update

* update

* update

* update

* update

Author: zhogu

.github/workflows/build-deploy-changes.yaml

@@ -171,13 +171,27 @@ jobs:
             --overwrite-existing
           kubelogin convert-kubeconfig -l azurecli
           kubectl config use-context ${{ secrets.KUBERNETES_CLUSTER }}
+            # Replace "webportal" with "webportal-dind" if "webportal" is changed
+            services_to_deploy="${{ steps.changes.outputs.folders }}"
+            if echo " $services_to_deploy " | grep -q " webportal "; then
+              tmp=""
+              for s in $services_to_deploy; do
+                [ "$s" = "webportal" ] && continue
+                [ "$s" = "webportal-dind" ] && continue
+                tmp="$tmp $s"
+              done
+              services_to_deploy="$tmp webportal-dind"
+              services_to_deploy=$(echo "$services_to_deploy" | xargs)
+            fi
+            echo "Final services to deploy: $services_to_deploy"
+
           echo "${{ secrets.PAI_CLUSTER_NAME }}" > cluster_id
-          echo "Stopping changed pai services \"${{ steps.changes.outputs.folders }}\" on ${{ secrets.PAI_CLUSTER_NAME }} ..."
-          $GITHUB_WORKSPACE/paictl.py service stop -n ${{ steps.changes.outputs.folders }} < cluster_id 
+          echo "Stopping changed pai services $services_to_deploy on ${{ secrets.PAI_CLUSTER_NAME }} ..."
+          $GITHUB_WORKSPACE/paictl.py service stop -n $services_to_deploy < cluster_id
           echo "Pushing config to cluster \"${{ secrets.PAI_CLUSTER_NAME }}\" ..."
-          $GITHUB_WORKSPACE/paictl.py config push -m service -p $GITHUB_WORKSPACE/config/cluster-configuration < cluster_id 
-          echo "Starting to update \"${{ steps.changes.outputs.folders }}\" on ${{ secrets.PAI_CLUSTER_NAME }} ..."
-          $GITHUB_WORKSPACE/paictl.py service start -n ${{ steps.changes.outputs.folders }} < cluster_id 
+          $GITHUB_WORKSPACE/paictl.py config push -m service -p $GITHUB_WORKSPACE/config/cluster-configuration < cluster_id
+          echo "Starting to update $services_to_deploy on ${{ secrets.PAI_CLUSTER_NAME }} ..."
+          $GITHUB_WORKSPACE/paictl.py service start -n $services_to_deploy < cluster_id
           kubectl get pod
           kubectl get service
 

src/webportal/src/app/job-submission/components/job-information.jsx

@@ -29,6 +29,7 @@ import { FormTextField } from './form-text-field';
 import { FormPage } from './form-page';
 import { FormSpinButton } from './form-spin-button';
 import { VirtualCluster } from './virtual-cluster';
+import { JobType } from './job-type';
 import Card from '../../components/card';
 import { JobBasicInfo } from '../models/job-basic-info';
 import { PROTOCOL_TOOLTIPS } from '../utils/constants';
@@ -37,7 +38,7 @@ const JOB_NAME_REGX = /^[A-Za-z0-9\-._~]+$/;
 
 export const JobInformation = React.memo(
   ({ jobInformation, onChange, advanceFlag }) => {
-    const { name, virtualCluster, jobRetryCount } = jobInformation;
+    const { name, virtualCluster, jobRetryCount, jobType } = jobInformation;
 
     const onChangeProp = useCallback(
       (type, value) => {
@@ -59,6 +60,11 @@ export const JobInformation = React.memo(
       [onChangeProp],
     );
 
+    const onJobTypeChange = useCallback(
+      jobType => onChangeProp('jobType', jobType),
+      [onChangeProp],
+    );
+
     const onRetryCountChange = useCallback(
       val => onChangeProp('jobRetryCount', val),
       [onChangeProp],
@@ -89,6 +95,10 @@ export const JobInformation = React.memo(
               onChange={onRetryCountChange}
             />
           )}
+          <JobType
+            onChange={onJobTypeChange}
+            jobType={jobType}
+          />
         </FormPage>
       </Card>
     );

src/webportal/src/app/job-submission/components/job-type.jsx

@@ -0,0 +1,51 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT License.
+
+import React, { useMemo, useCallback } from 'react';
+import { BasicSection } from './basic-section';
+import { Dropdown } from 'office-ui-fabric-react';
+import { FormShortSection } from './form-page';
+import PropTypes from 'prop-types';
+
+export const JobType = React.memo(props => {
+  const { onChange, jobType } = props;
+  const jobTypes = ['others', 'training', 'inference'];
+
+  const options = useMemo(
+    () =>
+      jobTypes.map((jobType, index) => {
+        return {
+          key: `jobType_${index}`,
+          text: jobType,
+        };
+      }),
+  );
+
+  const _onChange = useCallback(
+    (_, item) => {
+      if (onChange !== undefined) {
+        onChange(item.text);
+      }
+    },
+    [onChange],
+  );
+
+  const jobTypeIndex = options.findIndex(value => value.text === jobType);
+  return (
+    <BasicSection sectionLabel={'Job type'}>
+      <FormShortSection>
+        <Dropdown
+          placeholder='Select an option'
+          options={options}
+          onChange={_onChange}
+          selectedKey={jobTypeIndex === -1 ? "jobType_0" : `jobType_${jobTypeIndex}`}
+        />
+      </FormShortSection>
+    </BasicSection>
+  );
+});
+
+JobType.propTypes = {
+  onChange: PropTypes.func,
+  jobType: PropTypes.string,
+};

src/webportal/src/app/job-submission/job-submission-page.jsx

@@ -250,7 +250,7 @@ export const JobSubmissionPage = ({
             break;
           }
         }
-      } catch {} // ignore all exceptions here
+      } catch { } // ignore all exceptions here
       if (!isEmpty(defaultStorageConfig)) {
         const storagePlugin = {
           plugin: STORAGE_PLUGIN,
@@ -386,6 +386,33 @@ export const JobSubmissionPage = ({
       .catch(alert);
   }, [jobInformation.virtualCluster]);
 
+  useEffect(() => {
+    if (jobInformation.jobType === 'inference') {
+      const hasApiKey = parameters.find(param => param.key === 'API_KEY');
+      const hasInternalServerIp = parameters.find(param => param.key === 'INTERNAL_SERVER_IP');
+      const hasInternalServerPort = parameters.find(param => param.key === 'INTERNAL_SERVER_PORT');
+      if (!hasApiKey || !hasInternalServerIp || !hasInternalServerPort) {
+        const newParameters = [...parameters];
+        if (!hasInternalServerIp) {
+          // set fixed INTERNAL_SERVER_IP
+          newParameters.push({ key: 'INTERNAL_SERVER_IP', value: '$PAI_HOST_IP_taskrole_0' });
+        }
+        if (!hasInternalServerPort) {
+          // set fixed INTERNAL_SERVER_PORT
+          newParameters.push({ key: 'INTERNAL_SERVER_PORT', value: '$PAI_PORT_LIST_taskrole_0_http' });
+        }
+        if (!hasApiKey) {
+          {
+            // set a random generated api key
+            const randomKey = crypto.randomUUID();
+            newParameters.push({ key: 'API_KEY', value: randomKey });
+          }
+        }
+        setParameters(newParameters);
+      }
+    }
+  }, [jobInformation.jobType]);
+
   const onTemplateChange = useCallback((_, item) => {
     if (item.key === 'No') {
       return;

src/webportal/src/app/job-submission/models/job-basic-info.js

@@ -27,19 +27,21 @@ import { isEmpty, get } from 'lodash';
 
 export class JobBasicInfo {
   constructor(props) {
-    const { name, jobRetryCount, virtualCluster } = props;
+    const { name, jobRetryCount, virtualCluster, jobType } = props;
     this.name = name || '';
     this.jobRetryCount = jobRetryCount || 0;
     this.virtualCluster = virtualCluster || '';
+    this.jobType = jobType || '';
   }
 
   static fromProtocol(protocol) {
-    const { name, jobRetryCount } = protocol;
+    const { name, jobRetryCount, jobType } = protocol;
     const virtualCluster = get(protocol, 'defaults.virtualCluster', 'default');
     return new JobBasicInfo({
       name: name,
       jobRetryCount: jobRetryCount,
       virtualCluster: virtualCluster,
+      jobType: jobType,
     });
   }
 
@@ -59,6 +61,7 @@ export class JobBasicInfo {
       name: this.name,
       type: 'job',
       jobRetryCount: this.jobRetryCount,
+      jobType: this.jobType,
     };
   }
 }

src/webportal/src/app/job-submission/models/job-protocol.js

@@ -37,6 +37,7 @@ export class JobProtocol {
     const {
       name,
       jobRetryCount,
+      jobType,
       prerequisites,
       parameters,
       taskRoles,
@@ -53,6 +54,7 @@ export class JobProtocol {
     this.contributor = contributor || '';
     this.type = 'job';
     this.jobRetryCount = jobRetryCount || 0;
+    this.jobType = jobType || 'others';
     this.prerequisites = prerequisites || [];
     this.parameters = parameters || {};
     this.taskRoles = taskRoles || {};

src/webportal/src/app/job-submission/models/protocol-schema.js

@@ -173,6 +173,7 @@ export const jobProtocolSchema = Joi.object().keys({
   version: [Joi.string(), Joi.number()],
   contributor: Joi.string(),
   description: Joi.string(),
+  jobType: Joi.string().valid(['others', 'training', 'inference']).default('others'),
 
   prerequisites: Joi.array()
     .items(prerequisitesSchema)