lightdash 对于headless browser的使用

对于slack 的unfurl lightdash 使用了headless browser,以下说明下内部处理

参考图

此图来自官方文档,可以看出主要包含了调度以及slack 的

内部实现

实际上就是对于 unfurl 的图片处理部分,里边包含了cookie处理(安全)以及对于请求进行拦截(核心也是为了安全)同时还支持基于一些
配置对于页面的信息重新设置,整体使用上都是标准的puppeteer 操作,还是值得学习参考的

private async saveScreenshot({
        imageId,
        cookie,
        url,
        lightdashPage,
        chartType,
        organizationUuid,
        userUuid,
        gridWidth = undefined,
        resourceUuid = undefined,
        resourceName = undefined,
    }: {
        imageId: string;
        cookie: string;
        url: string;
        lightdashPage: LightdashPage;
        chartType?: string;
        organizationUuid?: string;
        userUuid: string;
        gridWidth?: number | undefined;
        resourceUuid?: string;
        resourceName?: string;
    }): Promise<Buffer | undefined> {
        if (this.lightdashConfig.headlessBrowser?.host === undefined) {
            this.logger.error(
                `Can't get screenshot if HEADLESS_BROWSER_HOST env variable is not defined`,
            );
            throw new Error(
                `Can't get screenshot if HEADLESS_BROWSER_HOST env variable is not defined`,
            );
        }
        const startTime = Date.now();
        let hasError = false;
 
        const isPuppeteerSetViewportDynamicallyEnabled =
            await isFeatureFlagEnabled(
                FeatureFlags.PuppeteerSetViewportDynamically,
                { userUuid, organizationUuid },
            );
        const isPuppeteerScrollElementIntoViewEnabled =
            await isFeatureFlagEnabled(
                FeatureFlags.PuppeteerScrollElementIntoView,
                { userUuid, organizationUuid },
            );
 
        return tracer.startActiveSpan(
            'UnfurlService.saveScreenshot',
            async (span) => {
                let browser;
 
                try {
                    const browserWSEndpoint = `ws://${
                        this.lightdashConfig.headlessBrowser?.host
                    }:${this.lightdashConfig.headlessBrowser?.port || 3001}`;
                    browser = await puppeteer.connect({
                        browserWSEndpoint,
                    });
 
                    const page = await browser.newPage();
                    const parsedUrl = new URL(url);
 
                    const cookieMatch = cookie.match(/connect\.sid=([^;]+)/); // Extract cookie value
                    if (!cookieMatch)
                        throw new Error('Invalid cookie provided');
                    const cookieValue = cookieMatch[1];
                    // Set cookie using `setCookie` instead of `setExtraHTTPHeaders` , otherwise this cookie will be leaked into other domains
                    await page.setCookie({
                        name: 'connect.sid',
                        value: cookieValue,
                        domain: parsedUrl.hostname, // Don't use ports here, cookies do not provide isolation by port
                        sameSite: 'Strict',
                    });
 
                    if (chartType === ChartType.BIG_NUMBER) {
                        await page.setViewport(bigNumberViewport);
                    } else {
                        await page.setViewport({
                            ...viewport,
                            width: gridWidth ?? viewport.width,
                        });
                    }
                    page.on('requestfailed', (request) => {
                        this.logger.warn(
                            `Headless browser request error - method: ${request.method()}, url: ${request.url()}, text: ${
                                request.failure()?.errorText
                            }`,
                        );
                    });
                    page.on('console', (msg) => {
                        const type = msg.type();
                        if (type === 'error') {
                            this.logger.warn(
                                `Headless browser console error - file: ${
                                    msg.location().url
                                }, text ${msg.text()} `,
                            );
                        }
                    });
                    /*
                    // This code can be used to block requests to external domains
                    // We disabled this so people can use images on markdown
                    await page.setRequestInterception(true);
                    await page.on('request', (request: HTTPRequest) => {
                        const requestUrl = request.url();
                        const cookie = request.headers()['cookie']
                        const parsedUrl = new URL(url);
                        // Only allow request to the same host
                        if (!requestUrl.includes(parsedUrl.origin)) {
                            request.abort();
                            return;
                        }
                        request.continue();
                    });
                    */
                    let chartRequests = 0;
                    let chartRequestErrors = 0;
 
                    page.on('response', (response) => {
                        const responseUrl = response.url();
                        const regexUrlToMatch =
                            lightdashPage === LightdashPage.EXPLORE
                                ? /\/saved\/[a-f0-9-]+\/results/
                                : /\/saved\/[a-f0-9-]+\/chart-and-results/; // NOTE: Chart endpoint in Dashboards is different
                        if (responseUrl.match(regexUrlToMatch)) {
                            chartRequests += 1;
                            response.buffer().then(
                                (buffer) => {
                                    const status = response.status();
                                    if (status >= 400) {
                                        this.logger.error(
                                            `Headless browser response error - url: ${responseUrl}, code: ${response.status()}, text: ${buffer}`,
                                        );
                                        chartRequestErrors += 1;
                                    }
                                },
                                (error) => {
                                    this.logger.error(
                                        `Headless browser response buffer error: ${error.message}`,
                                    );
                                    chartRequestErrors += 1;
                                },
                            );
                        }
                    });
                    let timeout = false;
                    try {
                        await page.goto(url, {
                            timeout: 150000, // Wait 2.5 mins for the page to load
                            waitUntil: 'networkidle0',
                        });
                    } catch (e) {
                        timeout = true;
                        this.logger.warn(
                            `Got a timeout when waiting for the page to load, returning current content`,
                        );
                    }
                    // Wait until the page is fully loaded
                    await page
                        .waitForSelector('.loading_chart', {
                            hidden: true,
                            timeout: 30000,
                        })
                        .catch(() => {
                            timeout = true;
                            this.logger.warn(
                                `Got a timeout when waiting for all charts to be loaded, returning current content`,
                            );
                        });
 
                    const path = `/tmp/${imageId}.png`;
                    let selector =
                        lightdashPage === LightdashPage.EXPLORE
                            ? `[data-testid="visualization"]`
                            : 'body';
                    if (
                        isPuppeteerSetViewportDynamicallyEnabled &&
                        lightdashPage === LightdashPage.DASHBOARD
                    ) {
                        selector = '.react-grid-layout';
                    }
 
                    const element = await page.waitForSelector(selector, {
                        timeout: 60000,
                    });
 
                    if (
                        isPuppeteerSetViewportDynamicallyEnabled &&
                        lightdashPage === LightdashPage.DASHBOARD
                    ) {
                        const fullPage = await page.$('.react-grid-layout');
                        const fullPageSize = await fullPage?.boundingBox();
                        await page.setViewport({
                            width: gridWidth ?? viewport.width,
                            height: fullPageSize?.height
                                ? parseInt(fullPageSize.height.toString(), 10)
                                : viewport.height,
                        });
                    }
 
                    if (!element) {
                        this.logger.warn(`Can't find element on page`);
                        return undefined;
                    }
 
                    const box = await element.boundingBox();
                    const pageMetrics = await page.metrics();
 
                    chartCounter.addCallback(async (result) => {
                        result.observe(chartRequests, {
                            errors: chartRequestErrors,
                            timeout,
                            organization_uuid: organizationUuid || 'undefined',
                        });
                    });
 
                    span.setAttributes({
                        'page.width': box?.width,
                        'page.height': box?.height,
                        'chart.requests.total': chartRequests,
                        'chart.requests.error': chartRequestErrors,
                        'page.metrics.task_duration': pageMetrics.TaskDuration,
                        'page.metrics.heap_size': pageMetrics.JSHeapUsedSize,
                        'page.metrics.total_size': pageMetrics.JSHeapTotalSize,
                        'page.type': lightdashPage,
                        url,
                        chartType: chartType || 'undefined',
                        organization_uuid: organizationUuid || 'undefined',
                        'page.metrics.event_listeners':
                            pageMetrics.JSEventListeners,
                        timeout,
                    });
 
                    if (this.lightdashConfig.scheduler.screenshotTimeout) {
                        await new Promise((resolve) => {
                            setTimeout(
                                resolve,
                                this.lightdashConfig.scheduler
                                    .screenshotTimeout,
                            );
                        });
                    }
                    const imageBuffer = await element.screenshot({
                        path,
                        ...(isPuppeteerScrollElementIntoViewEnabled &&
                        lightdashPage === LightdashPage.DASHBOARD
                            ? {
                                  scrollIntoView: true,
                              }
                            : {}),
                    });
                    return imageBuffer;
                } catch (e) {
                    Sentry.captureException(e);
                    hasError = true;
                    span.recordException(e);
                    span.setAttributes({
                        'page.type': lightdashPage,
                        url,
                        chartType: chartType || 'undefined',
                        organization_uuid: organizationUuid || 'undefined',
                        uuid: resourceUuid ?? 'undefined',
                        title: resourceName ?? 'undefined',
                        is_viewport_dynamically_enabled: `${isPuppeteerSetViewportDynamicallyEnabled}`,
                        is_scroll_into_view_enabled: `${isPuppeteerScrollElementIntoViewEnabled}`,
                        custom_width: `${gridWidth}`,
                    });
                    span.setStatus({
                        code: SpanStatusCode.ERROR,
                    });
 
                    this.logger.error(
                        `Unable to fetch screenshots for scheduler with url ${url}, of type: ${lightdashPage}. Message: ${e.message}`,
                    );
                    throw e;
                } finally {
                    if (browser) await browser.close();
 
                    span.end();
 
                    const executionTime = Date.now() - startTime;
                    this.logger.info(
                        `UnfurlService saveScreenshot took ${executionTime} ms`,
                    );
                    taskDurationHistogram.record(executionTime, {
                        error: hasError,
                    });
                }
            },
        );
    }

说明

日常中需要进行快照的场景还是不少的,基于无头浏览器是一个很不错的选择,成本相对比,开发也比较快速,browserless 是一个很不错的工具

参考资料

https://docs.lightdash.com/self-host/customize-deployment/enable-headless-browser-for-lightdash/
https://pptr.dev/
https://github.com/browserless/browserless

posted on 2024-04-18 11:48  荣锋亮  阅读(10)  评论(0编辑  收藏  举报

导航