amlEnvironmentService.ts 5.78 KB
Newer Older
SparkSnail's avatar
SparkSnail committed
1
2
3
4
5
// Copyright (c) Microsoft Corporation.
// Licensed under the MIT license.

'use strict';

6
import { EventEmitter } from "events";
SparkSnail's avatar
SparkSnail committed
7
8
9
10
11
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';
12
import { getExperimentRootDir } from '../../../common/utils';
SparkSnail's avatar
SparkSnail committed
13
14
import { TrialConfigMetadataKey } from '../../common/trialConfigMetadataKey';
import { validateCodeDir } from '../../common/util';
15
16
import { AMLClient } from '../aml/amlClient';
import { AMLClusterConfig, AMLEnvironmentInformation, AMLTrialConfig } from '../aml/amlConfig';
SparkSnail's avatar
SparkSnail committed
17
18
import { AMLCommandChannel } from '../channels/amlCommandChannel';
import { CommandChannel } from "../commandChannel";
19
import { EnvironmentInformation, EnvironmentService } from '../environment';
SparkSnail's avatar
SparkSnail committed
20
21
22
23
24
25
26


/**
 * Collector PAI jobs info from PAI cluster, and update pai job status locally
 */
@component.Singleton
export class AMLEnvironmentService extends EnvironmentService {
27

SparkSnail's avatar
SparkSnail committed
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
    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;
    }

44
    public createCommandChannel(commandEmitter: EventEmitter): CommandChannel {
SparkSnail's avatar
SparkSnail committed
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
        return new AMLCommandChannel(commandEmitter);
    }

    public createEnviornmentInfomation(envId: string, envName: string): EnvironmentInformation {
        return new AMLEnvironmentInformation(envId, envName);
    }

    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;
76
            if (!amlClient) {
77
                return Promise.reject('AML client not initialized!');
SparkSnail's avatar
SparkSnail committed
78
            }
79
80
            const newStatus = await amlClient.updateStatus(environment.status);
            switch (newStatus.toUpperCase()) {
SparkSnail's avatar
SparkSnail committed
81
82
                case 'WAITING':
                case 'QUEUED':
83
84
85
86
                    environment.setStatus('WAITING');
                    break;
                case 'RUNNING':
                    environment.setStatus('RUNNING');
SparkSnail's avatar
SparkSnail committed
87
88
89
                    break;
                case 'COMPLETED':
                case 'SUCCEEDED':
90
                    environment.setStatus('SUCCEEDED');
SparkSnail's avatar
SparkSnail committed
91
92
                    break;
                case 'FAILED':
93
94
                    environment.setStatus('FAILED');
                    return Promise.reject(`AML: job ${environment.envId} is failed!`);
SparkSnail's avatar
SparkSnail committed
95
96
                case 'STOPPED':
                case 'STOPPING':
97
                    environment.setStatus('USER_CANCELED');
SparkSnail's avatar
SparkSnail committed
98
99
                    break;
                default:
100
                    environment.setStatus('UNKNOWN');
SparkSnail's avatar
SparkSnail committed
101
102
103
104
105
106
107
108
109
110
111
112
113
114
            }
        });
    }

    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");
        environment.command = `import os\nos.system('${amlEnvironment.command}')`;
115
116
        environment.useActiveGpu = this.amlClusterConfig.useActiveGpu;
        environment.maxTrialNumberPerGpu = this.amlClusterConfig.maxTrialNumPerGpu;
117
        await fs.promises.writeFile(path.join(environmentLocalTempFolder, 'nni_script.py'), amlEnvironment.command, { encoding: 'utf8' });
SparkSnail's avatar
SparkSnail committed
118
119
120
121
122
        const amlClient = new AMLClient(
            this.amlClusterConfig.subscriptionId,
            this.amlClusterConfig.resourceGroup,
            this.amlClusterConfig.workspaceName,
            this.experimentId,
123
            this.amlClusterConfig.computeTarget,
SparkSnail's avatar
SparkSnail committed
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
            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();
    }
}