toolkit/packages/artifact/src/internal/download-http-client.ts
Konrad Pabjan c861dd8859
Retry all http calls for artifact upload and download (#675)
* Retry all http calls for artifact upload and download

* Extra debug information

* Fix lint

* Always read response body

* PR Feedback

* Change error message if patch call fails

* Add exponential backoff when retrying

* Rework tests and add diagnostic info if exception thrown

* Fix lint

* fix lint error for real this time

* PR cleanup

* 0.5.0 @actions/artifact release

* Display diagnostic info if non-retryable code is hit
2020-12-18 21:40:50 +01:00

359 lines
13 KiB
TypeScript

import * as fs from 'fs'
import * as core from '@actions/core'
import * as zlib from 'zlib'
import {
getArtifactUrl,
getDownloadHeaders,
isSuccessStatusCode,
isRetryableStatusCode,
isThrottledStatusCode,
getExponentialRetryTimeInMilliseconds,
tryGetRetryAfterValueTimeInMilliseconds,
displayHttpDiagnostics,
getFileSize,
rmFile,
sleep
} from './utils'
import {URL} from 'url'
import {StatusReporter} from './status-reporter'
import {performance} from 'perf_hooks'
import {ListArtifactsResponse, QueryArtifactResponse} from './contracts'
import {IHttpClientResponse} from '@actions/http-client/interfaces'
import {HttpManager} from './http-manager'
import {DownloadItem} from './download-specification'
import {getDownloadFileConcurrency, getRetryLimit} from './config-variables'
import {IncomingHttpHeaders} from 'http'
import {retryHttpClientRequest} from './requestUtils'
export class DownloadHttpClient {
// http manager is used for concurrent connections when downloading multiple files at once
private downloadHttpManager: HttpManager
private statusReporter: StatusReporter
constructor() {
this.downloadHttpManager = new HttpManager(
getDownloadFileConcurrency(),
'@actions/artifact-download'
)
// downloads are usually significantly faster than uploads so display status information every second
this.statusReporter = new StatusReporter(1000)
}
/**
* Gets a list of all artifacts that are in a specific container
*/
async listArtifacts(): Promise<ListArtifactsResponse> {
const artifactUrl = getArtifactUrl()
// use the first client from the httpManager, `keep-alive` is not used so the connection will close immediately
const client = this.downloadHttpManager.getClient(0)
const headers = getDownloadHeaders('application/json')
const response = await retryHttpClientRequest('List Artifacts', async () =>
client.get(artifactUrl, headers)
)
const body: string = await response.readBody()
return JSON.parse(body)
}
/**
* Fetches a set of container items that describe the contents of an artifact
* @param artifactName the name of the artifact
* @param containerUrl the artifact container URL for the run
*/
async getContainerItems(
artifactName: string,
containerUrl: string
): Promise<QueryArtifactResponse> {
// the itemPath search parameter controls which containers will be returned
const resourceUrl = new URL(containerUrl)
resourceUrl.searchParams.append('itemPath', artifactName)
// use the first client from the httpManager, `keep-alive` is not used so the connection will close immediately
const client = this.downloadHttpManager.getClient(0)
const headers = getDownloadHeaders('application/json')
const response = await retryHttpClientRequest(
'Get Container Items',
async () => client.get(resourceUrl.toString(), headers)
)
const body: string = await response.readBody()
return JSON.parse(body)
}
/**
* Concurrently downloads all the files that are part of an artifact
* @param downloadItems information about what items to download and where to save them
*/
async downloadSingleArtifact(downloadItems: DownloadItem[]): Promise<void> {
const DOWNLOAD_CONCURRENCY = getDownloadFileConcurrency()
// limit the number of files downloaded at a single time
core.debug(`Download file concurrency is set to ${DOWNLOAD_CONCURRENCY}`)
const parallelDownloads = [...new Array(DOWNLOAD_CONCURRENCY).keys()]
let currentFile = 0
let downloadedFiles = 0
core.info(
`Total number of files that will be downloaded: ${downloadItems.length}`
)
this.statusReporter.setTotalNumberOfFilesToProcess(downloadItems.length)
this.statusReporter.start()
await Promise.all(
parallelDownloads.map(async index => {
while (currentFile < downloadItems.length) {
const currentFileToDownload = downloadItems[currentFile]
currentFile += 1
const startTime = performance.now()
await this.downloadIndividualFile(
index,
currentFileToDownload.sourceLocation,
currentFileToDownload.targetPath
)
if (core.isDebug()) {
core.debug(
`File: ${++downloadedFiles}/${downloadItems.length}. ${
currentFileToDownload.targetPath
} took ${(performance.now() - startTime).toFixed(
3
)} milliseconds to finish downloading`
)
}
this.statusReporter.incrementProcessedCount()
}
})
)
.catch(error => {
throw new Error(`Unable to download the artifact: ${error}`)
})
.finally(() => {
this.statusReporter.stop()
// safety dispose all connections
this.downloadHttpManager.disposeAndReplaceAllClients()
})
}
/**
* Downloads an individual file
* @param httpClientIndex the index of the http client that is used to make all of the calls
* @param artifactLocation origin location where a file will be downloaded from
* @param downloadPath destination location for the file being downloaded
*/
private async downloadIndividualFile(
httpClientIndex: number,
artifactLocation: string,
downloadPath: string
): Promise<void> {
let retryCount = 0
const retryLimit = getRetryLimit()
let destinationStream = fs.createWriteStream(downloadPath)
const headers = getDownloadHeaders('application/json', true, true)
// a single GET request is used to download a file
const makeDownloadRequest = async (): Promise<IHttpClientResponse> => {
const client = this.downloadHttpManager.getClient(httpClientIndex)
return await client.get(artifactLocation, headers)
}
// check the response headers to determine if the file was compressed using gzip
const isGzip = (incomingHeaders: IncomingHttpHeaders): boolean => {
return (
'content-encoding' in incomingHeaders &&
incomingHeaders['content-encoding'] === 'gzip'
)
}
// Increments the current retry count and then checks if the retry limit has been reached
// If there have been too many retries, fail so the download stops. If there is a retryAfterValue value provided,
// it will be used
const backOff = async (retryAfterValue?: number): Promise<void> => {
retryCount++
if (retryCount > retryLimit) {
return Promise.reject(
new Error(
`Retry limit has been reached. Unable to download ${artifactLocation}`
)
)
} else {
this.downloadHttpManager.disposeAndReplaceClient(httpClientIndex)
if (retryAfterValue) {
// Back off by waiting the specified time denoted by the retry-after header
core.info(
`Backoff due to too many requests, retry #${retryCount}. Waiting for ${retryAfterValue} milliseconds before continuing the download`
)
await sleep(retryAfterValue)
} else {
// Back off using an exponential value that depends on the retry count
const backoffTime = getExponentialRetryTimeInMilliseconds(retryCount)
core.info(
`Exponential backoff for retry #${retryCount}. Waiting for ${backoffTime} milliseconds before continuing the download`
)
await sleep(backoffTime)
}
core.info(
`Finished backoff for retry #${retryCount}, continuing with download`
)
}
}
const isAllBytesReceived = (
expected?: string,
received?: number
): boolean => {
// be lenient, if any input is missing, assume success, i.e. not truncated
if (
!expected ||
!received ||
process.env['ACTIONS_ARTIFACT_SKIP_DOWNLOAD_VALIDATION']
) {
core.info('Skipping download validation.')
return true
}
return parseInt(expected) === received
}
const resetDestinationStream = async (
fileDownloadPath: string
): Promise<void> => {
destinationStream.close()
await rmFile(fileDownloadPath)
destinationStream = fs.createWriteStream(fileDownloadPath)
}
// keep trying to download a file until a retry limit has been reached
while (retryCount <= retryLimit) {
let response: IHttpClientResponse
try {
response = await makeDownloadRequest()
if (core.isDebug()) {
displayHttpDiagnostics(response)
}
} catch (error) {
// if an error is caught, it is usually indicative of a timeout so retry the download
core.info('An error occurred while attempting to download a file')
// eslint-disable-next-line no-console
console.log(error)
// increment the retryCount and use exponential backoff to wait before making the next request
await backOff()
continue
}
let forceRetry = false
if (isSuccessStatusCode(response.message.statusCode)) {
// The body contains the contents of the file however calling response.readBody() causes all the content to be converted to a string
// which can cause some gzip encoded data to be lost
// Instead of using response.readBody(), response.message is a readableStream that can be directly used to get the raw body contents
try {
const isGzipped = isGzip(response.message.headers)
await this.pipeResponseToFile(response, destinationStream, isGzipped)
if (
isGzipped ||
isAllBytesReceived(
response.message.headers['content-length'],
await getFileSize(downloadPath)
)
) {
return
} else {
forceRetry = true
}
} catch (error) {
// retry on error, most likely streams were corrupted
forceRetry = true
}
}
if (forceRetry || isRetryableStatusCode(response.message.statusCode)) {
core.info(
`A ${response.message.statusCode} response code has been received while attempting to download an artifact`
)
resetDestinationStream(downloadPath)
// if a throttled status code is received, try to get the retryAfter header value, else differ to standard exponential backoff
isThrottledStatusCode(response.message.statusCode)
? await backOff(
tryGetRetryAfterValueTimeInMilliseconds(response.message.headers)
)
: await backOff()
} else {
// Some unexpected response code, fail immediately and stop the download
displayHttpDiagnostics(response)
return Promise.reject(
new Error(
`Unexpected http ${response.message.statusCode} during download for ${artifactLocation}`
)
)
}
}
}
/**
* Pipes the response from downloading an individual file to the appropriate destination stream while decoding gzip content if necessary
* @param response the http response received when downloading a file
* @param destinationStream the stream where the file should be written to
* @param isGzip a boolean denoting if the content is compressed using gzip and if we need to decode it
*/
async pipeResponseToFile(
response: IHttpClientResponse,
destinationStream: fs.WriteStream,
isGzip: boolean
): Promise<void> {
await new Promise((resolve, reject) => {
if (isGzip) {
const gunzip = zlib.createGunzip()
response.message
.on('error', error => {
core.error(
`An error occurred while attempting to read the response stream`
)
gunzip.close()
destinationStream.close()
reject(error)
})
.pipe(gunzip)
.on('error', error => {
core.error(
`An error occurred while attempting to decompress the response stream`
)
destinationStream.close()
reject(error)
})
.pipe(destinationStream)
.on('close', () => {
resolve()
})
.on('error', error => {
core.error(
`An error occurred while writing a downloaded file to ${destinationStream.path}`
)
reject(error)
})
} else {
response.message
.on('error', error => {
core.error(
`An error occurred while attempting to read the response stream`
)
destinationStream.close()
reject(error)
})
.pipe(destinationStream)
.on('close', () => {
resolve()
})
.on('error', error => {
core.error(
`An error occurred while writing a downloaded file to ${destinationStream.path}`
)
reject(error)
})
}
})
return
}
}