amlEnvironmentService.ts 5.83 KB
Newer Older
SparkSnail's avatar
SparkSnail committed
1
2
3
4
5
6
7
8
9
10
// Copyright (c) Microsoft Corporation.
// Licensed under the MIT license.

'use strict';

import * as fs from 'fs';
import * as path from 'path';
import * as component from '../../../common/component';
import { getExperimentId } from '../../../common/experimentStartupInfo';
import { getLogger, Logger } from '../../../common/log';
11
import { getExperimentRootDir } from '../../../common/utils';
SparkSnail's avatar
SparkSnail committed
12
13
import { TrialConfigMetadataKey } from '../../common/trialConfigMetadataKey';
import { validateCodeDir } from '../../common/util';
14
15
import { AMLClient } from '../aml/amlClient';
import { AMLClusterConfig, AMLEnvironmentInformation, AMLTrialConfig } from '../aml/amlConfig';
16
import { EnvironmentInformation, EnvironmentService } from '../environment';
17
18
import { EventEmitter } from "events";
import { AMLCommandChannel } from '../channels/amlCommandChannel';
SparkSnail's avatar
SparkSnail committed
19
20
21


/**
22
 * Collector AML jobs info from AML cluster, and update aml job status locally
SparkSnail's avatar
SparkSnail committed
23
24
25
 */
@component.Singleton
export class AMLEnvironmentService extends EnvironmentService {
26

SparkSnail's avatar
SparkSnail committed
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
    private readonly log: Logger = getLogger();
    public amlClusterConfig: AMLClusterConfig | undefined;
    public amlTrialConfig: AMLTrialConfig | undefined;
    private experimentId: string;
    private experimentRootDir: string;

    constructor() {
        super();
        this.experimentId = getExperimentId();
        this.experimentRootDir = getExperimentRootDir();
    }

    public get hasStorageService(): boolean {
        return false;
    }

43
44
    public initCommandChannel(eventEmitter: EventEmitter): void {
        this.commandChannel = new AMLCommandChannel(eventEmitter);
SparkSnail's avatar
SparkSnail committed
45
46
    }

J-shang's avatar
J-shang committed
47
    public createEnvironmentInformation(envId: string, envName: string): EnvironmentInformation {
SparkSnail's avatar
SparkSnail committed
48
49
50
        return new AMLEnvironmentInformation(envId, envName);
    }

51
52
53
54
    public get getName(): string {
        return 'aml';
    }

SparkSnail's avatar
SparkSnail committed
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
    public async config(key: string, value: string): Promise<void> {
        switch (key) {
            case TrialConfigMetadataKey.AML_CLUSTER_CONFIG:
                this.amlClusterConfig = <AMLClusterConfig>JSON.parse(value);
                break;

            case TrialConfigMetadataKey.TRIAL_CONFIG: {
                if (this.amlClusterConfig === undefined) {
                    this.log.error('aml cluster config is not initialized');
                    break;
                }
                this.amlTrialConfig = <AMLTrialConfig>JSON.parse(value);
                // Validate to make sure codeDir doesn't have too many files
                await validateCodeDir(this.amlTrialConfig.codeDir);
                break;
            }
            default:
                this.log.debug(`AML not proccessed metadata key: '${key}', value: '${value}'`);
        }
    }

    public async refreshEnvironmentsStatus(environments: EnvironmentInformation[]): Promise<void> {
        environments.forEach(async (environment) => {
            const amlClient = (environment as AMLEnvironmentInformation).amlClient;
79
            if (!amlClient) {
80
                return Promise.reject('AML client not initialized!');
SparkSnail's avatar
SparkSnail committed
81
            }
82
83
            const newStatus = await amlClient.updateStatus(environment.status);
            switch (newStatus.toUpperCase()) {
SparkSnail's avatar
SparkSnail committed
84
85
                case 'WAITING':
                case 'QUEUED':
86
87
88
89
                    environment.setStatus('WAITING');
                    break;
                case 'RUNNING':
                    environment.setStatus('RUNNING');
SparkSnail's avatar
SparkSnail committed
90
91
92
                    break;
                case 'COMPLETED':
                case 'SUCCEEDED':
93
                    environment.setStatus('SUCCEEDED');
SparkSnail's avatar
SparkSnail committed
94
95
                    break;
                case 'FAILED':
96
97
                    environment.setStatus('FAILED');
                    return Promise.reject(`AML: job ${environment.envId} is failed!`);
SparkSnail's avatar
SparkSnail committed
98
99
                case 'STOPPED':
                case 'STOPPING':
100
                    environment.setStatus('USER_CANCELED');
SparkSnail's avatar
SparkSnail committed
101
102
                    break;
                default:
103
                    environment.setStatus('UNKNOWN');
SparkSnail's avatar
SparkSnail committed
104
105
106
107
108
109
110
111
112
113
114
115
116
            }
        });
    }

    public async startEnvironment(environment: EnvironmentInformation): Promise<void> {
        if (this.amlClusterConfig === undefined) {
            throw new Error('AML Cluster config is not initialized');
        }
        if (this.amlTrialConfig === undefined) {
            throw new Error('AML trial config is not initialized');
        }
        const amlEnvironment: AMLEnvironmentInformation = environment as AMLEnvironmentInformation;
        const environmentLocalTempFolder = path.join(this.experimentRootDir, this.experimentId, "environment-temp");
117
        environment.command = `import os\nos.system('mv envs outputs/envs && cd outputs && ${amlEnvironment.command}')`;
118
119
        environment.useActiveGpu = this.amlClusterConfig.useActiveGpu;
        environment.maxTrialNumberPerGpu = this.amlClusterConfig.maxTrialNumPerGpu;
120
        await fs.promises.writeFile(path.join(environmentLocalTempFolder, 'nni_script.py'), amlEnvironment.command, { encoding: 'utf8' });
SparkSnail's avatar
SparkSnail committed
121
122
123
124
125
        const amlClient = new AMLClient(
            this.amlClusterConfig.subscriptionId,
            this.amlClusterConfig.resourceGroup,
            this.amlClusterConfig.workspaceName,
            this.experimentId,
126
            this.amlClusterConfig.computeTarget,
SparkSnail's avatar
SparkSnail committed
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
            this.amlTrialConfig.image,
            'nni_script.py',
            environmentLocalTempFolder
        );
        amlEnvironment.id = await amlClient.submit();
        amlEnvironment.trackingUrl = await amlClient.getTrackingUrl();
        amlEnvironment.amlClient = amlClient;
    }

    public async stopEnvironment(environment: EnvironmentInformation): Promise<void> {
        const amlEnvironment: AMLEnvironmentInformation = environment as AMLEnvironmentInformation;
        const amlClient = amlEnvironment.amlClient;
        if (!amlClient) {
            throw new Error('AML client not initialized!');
        }
        amlClient.stop();
    }
}