1
0
Fork 0

Use zlib for compression

Link-/blobcache-spike
Bassem Dghaidi 2024-06-13 03:16:59 -07:00 committed by GitHub
parent 9e63a77e7a
commit 5e5faf73fc
8 changed files with 2632 additions and 48 deletions

View File

@ -2,10 +2,14 @@ import * as core from '@actions/core'
import * as path from 'path'
import {saveCache} from '../src/cache'
import * as cacheHttpClient from '../src/internal/cacheHttpClient'
import * as cacheTwirpClient from '../src/internal/cacheTwirpClient'
import {GetCacheBlobUploadURLResponse} from '../src/generated/results/api/v1/blobcache'
import {BlobCacheServiceClientJSON} from '../src/generated/results/api/v1/blobcache.twirp'
import * as cacheUtils from '../src/internal/cacheUtils'
import {CacheFilename, CompressionMethod} from '../src/internal/constants'
import * as tar from '../src/internal/tar'
import {TypedResponse} from '@actions/http-client/lib/interfaces'
import * as uploadCache from '../src/internal/v2/upload-cache'
import {
ReserveCacheResponse,
ITypedResponseWithError
@ -327,3 +331,74 @@ test('save with non existing path should not save cache', async () => {
`Path Validation Error: Path(s) specified in the action for caching do(es) not exist, hence no cache is being saved.`
)
})
test('throwaway test', async () => {
const filePath = 'node_modules'
const primaryKey = 'Linux-node-bb828da54c148048dd17899ba9fda624811cfb43'
const cachePaths = [path.resolve(filePath)]
const cacheSignedURL = 'https://container.blob.core.windows.net/cache/${primaryKey}?sig=1234'
const getCacheBlobUploadURL: GetCacheBlobUploadURLResponse = {
urls: [
{
key: primaryKey,
url: cacheSignedURL,
},
]
}
const cacheId = 4
const reserveCacheMock = jest
.spyOn(cacheHttpClient, 'reserveCache')
.mockImplementation(async () => {
const response: TypedResponse<ReserveCacheResponse> = {
statusCode: 500,
result: {cacheId},
headers: {}
}
return response
})
const getCacheBlobUploadURLMock = jest
.spyOn(BlobCacheServiceClientJSON.prototype, 'GetCacheBlobUploadURL')
.mockResolvedValue(getCacheBlobUploadURL)
const uploadCacheMock = jest
.spyOn(uploadCache, 'UploadCacheFile')
.mockImplementation(async () => {
return {
status: 200
}
})
const createTarMock = jest.spyOn(tar, 'createTar')
const saveCacheMock = jest.spyOn(cacheHttpClient, 'saveCache')
const compression = CompressionMethod.Zstd
const getCompressionMock = jest
.spyOn(cacheUtils, 'getCompressionMethod')
.mockReturnValue(Promise.resolve(compression))
await uploadCache.UploadCacheFile(getCacheBlobUploadURL, cachePaths[0])
await saveCache([filePath], primaryKey)
expect(reserveCacheMock).toHaveBeenCalledTimes(1)
expect(reserveCacheMock).toHaveBeenCalledWith(primaryKey, [filePath], {
cacheSize: undefined,
compressionMethod: compression,
enableCrossOsArchive: false
})
expect (getCacheBlobUploadURLMock).toHaveBeenCalledTimes(1)
const archiveFolder = '/foo/bar'
const archiveFile = path.join(archiveFolder, CacheFilename.Zstd)
expect(createTarMock).toHaveBeenCalledTimes(1)
expect(createTarMock).toHaveBeenCalledWith(
archiveFolder,
cachePaths,
compression
)
expect(uploadCacheMock).toHaveBeenCalledTimes(2)
expect(saveCacheMock).toHaveBeenCalledTimes(1)
expect(saveCacheMock).toHaveBeenCalledWith(cacheId, archiveFile, undefined)
expect(getCompressionMock).toHaveBeenCalledTimes(1)
})

2357
packages/cache/package-lock.json generated vendored

File diff suppressed because it is too large Load Diff

View File

@ -38,6 +38,7 @@
},
"dependencies": {
"@actions/core": "^1.10.0",
"@actions/artifact": "^2.1.7",
"@actions/exec": "^1.0.1",
"@actions/glob": "^0.1.0",
"@actions/http-client": "^2.1.1",

View File

@ -1,13 +1,18 @@
import * as core from '@actions/core'
import * as path from 'path'
import * as utils from './internal/cacheUtils'
import {CacheUrl} from './internal/constants'
import {CacheServiceVersion, CacheUrl} from './internal/constants'
import * as cacheHttpClient from './internal/cacheHttpClient'
import * as cacheTwirpClient from './internal/cacheTwirpClient'
import {createTar, extractTar, listTar} from './internal/tar'
import {DownloadOptions, UploadOptions} from './options'
import {GetCacheBlobUploadURLRequest, GetCacheBlobUploadURLResponse} from './generated/results/api/v1/blobcache'
import {UploadCache} from './internal/v2/upload/upload-cache'
import {UploadCacheStream} from './internal/v2/upload-cache'
import {
UploadZipSpecification,
getUploadZipSpecification
} from '@actions/artifact/lib/internal/upload/upload-zip-specification'
import {createZipUploadStream} from '@actions/artifact/lib/internal/upload/zip'
export class ValidationError extends Error {
constructor(message: string) {
@ -174,17 +179,23 @@ export async function saveCache(
): Promise<number> {
checkPaths(paths)
checkKey(key)
// TODO: REMOVE ME
// Making a call to the service
const twirpClient = cacheTwirpClient.internalBlobCacheTwirpClient()
const getSignedUploadURL: GetCacheBlobUploadURLRequest = {
organization: "github",
keys: [key],
console.debug(`Cache Service Version: ${CacheServiceVersion}`)
switch (CacheServiceVersion) {
case "v2":
return await saveCachev1(paths, key, options, enableCrossOsArchive)
case "v1":
default:
return await saveCachev2(paths, key, options, enableCrossOsArchive)
}
const signedUploadURL: GetCacheBlobUploadURLResponse = await twirpClient.GetCacheBlobUploadURL(getSignedUploadURL)
core.info(`GetCacheBlobUploadURLResponse: ${JSON.stringify(signedUploadURL)}`)
}
async function saveCachev1(
paths: string[],
key: string,
options?: UploadOptions,
enableCrossOsArchive = false
): Promise<number> {
const compressionMethod = await utils.getCompressionMethod()
let cacheId = -1
@ -224,15 +235,6 @@ export async function saveCache(
)
}
// Cache v2 upload
// inputs:
// - getSignedUploadURL
// - archivePath
core.info(`Saving Cache v2: ${archivePath}`)
await UploadCache(signedUploadURL, archivePath)
core.debug('Reserving Cache')
const reserveCacheResponse = await cacheHttpClient.reserveCache(
key,
@ -281,3 +283,47 @@ export async function saveCache(
return cacheId
}
async function saveCachev2(
paths: string[],
key: string,
options?: UploadOptions,
enableCrossOsArchive = false
): Promise<number> {
const twirpClient = cacheTwirpClient.internalBlobCacheTwirpClient()
const getSignedUploadURL: GetCacheBlobUploadURLRequest = {
organization: "github",
keys: [key],
}
const signedUploadURL: GetCacheBlobUploadURLResponse = await twirpClient.GetCacheBlobUploadURL(getSignedUploadURL)
core.info(`GetCacheBlobUploadURLResponse: ${JSON.stringify(signedUploadURL)}`)
// Archive
// We're going to handle 1 path fow now. This needs to be fixed to handle all
// paths passed in.
const rootDir = path.dirname(paths[0])
const zipSpecs: UploadZipSpecification[] = getUploadZipSpecification(paths, rootDir)
if (zipSpecs.length === 0) {
throw new Error(
`Error with zip specs: ${zipSpecs.flatMap(s => (s.sourcePath ? [s.sourcePath] : [])).join(', ')}`
)
}
// 0: No compression
// 1: Best speed
// 6: Default compression (same as GNU Gzip)
// 9: Best compression Higher levels will result in better compression, but will take longer to complete. For large files that are not easily compressed, a value of 0 is recommended for significantly faster uploads.
const zipUploadStream = await createZipUploadStream(
zipSpecs,
6
)
// Cache v2 upload
// inputs:
// - getSignedUploadURL
// - archivePath
core.info(`Saving Cache v2: ${paths[0]}`)
await UploadCacheStream(signedUploadURL.urls[0].url, zipUploadStream)
return 0
}

View File

@ -37,5 +37,6 @@ export const TarFilename = 'cache.tar'
export const ManifestFilename = 'manifest.txt'
// Cache URL
export const CacheUrl = `${process.env['ACTIONS_CACHE_URL_NEXT']}`
// Cache Service Metadata
export const CacheUrl = `${process.env['ACTIONS_CACHE_URL_NEXT']} || ${process.env['ACTIONS_CACHE_URL']}`
export const CacheServiceVersion = `${process.env['ACTIONS_CACHE_URL_NEXT']} ? 'v2' : 'v1'`

View File

@ -0,0 +1,130 @@
import * as core from '@actions/core'
import {GetCacheBlobUploadURLResponse} from '../../generated/results/api/v1/blobcache'
import {ZipUploadStream} from '@actions/artifact/lib/internal/upload/zip'
import {NetworkError} from '@actions/artifact/'
import {TransferProgressEvent} from '@azure/core-http'
import * as stream from 'stream'
import * as crypto from 'crypto'
import {
BlobClient,
BlockBlobClient,
BlockBlobUploadStreamOptions,
BlockBlobParallelUploadOptions
} from '@azure/storage-blob'
export async function UploadCacheStream(
signedUploadURL: string,
zipUploadStream: ZipUploadStream
): Promise<{}> {
let uploadByteCount = 0
let lastProgressTime = Date.now()
let timeoutId: NodeJS.Timeout | undefined
const chunkTimer = (timeout: number): NodeJS.Timeout => {
// clear the previous timeout
if (timeoutId) {
clearTimeout(timeoutId)
}
timeoutId = setTimeout(() => {
const now = Date.now()
// if there's been more than 30 seconds since the
// last progress event, then we'll consider the upload stalled
if (now - lastProgressTime > timeout) {
throw new Error('Upload progress stalled.')
}
}, timeout)
return timeoutId
}
const maxConcurrency = 32
const bufferSize = 8 * 1024 * 1024 // 8 MB Chunks
const blobClient = new BlobClient(signedUploadURL)
const blockBlobClient = blobClient.getBlockBlobClient()
const timeoutDuration = 300000 // 30 seconds
core.debug(
`Uploading cache zip to blob storage with maxConcurrency: ${maxConcurrency}, bufferSize: ${bufferSize}`
)
const uploadCallback = (progress: TransferProgressEvent): void => {
core.info(`Uploaded bytes ${progress.loadedBytes}`)
uploadByteCount = progress.loadedBytes
chunkTimer(timeoutDuration)
lastProgressTime = Date.now()
}
const options: BlockBlobUploadStreamOptions = {
blobHTTPHeaders: {blobContentType: 'zip'},
onProgress: uploadCallback
}
let sha256Hash: string | undefined = undefined
const uploadStream = new stream.PassThrough()
const hashStream = crypto.createHash('sha256')
zipUploadStream.pipe(uploadStream) // This stream is used for the upload
zipUploadStream.pipe(hashStream).setEncoding('hex') // This stream is used to compute a hash of the zip content that gets used. Integrity check
core.info('Beginning upload of cache to blob storage')
try {
// Start the chunk timer
timeoutId = chunkTimer(timeoutDuration)
await blockBlobClient.uploadStream(
uploadStream,
bufferSize,
maxConcurrency,
options
)
} catch (error) {
if (NetworkError.isNetworkErrorCode(error?.code)) {
throw new NetworkError(error?.code)
}
throw error
} finally {
// clear the timeout whether or not the upload completes
if (timeoutId) {
clearTimeout(timeoutId)
}
}
core.info('Finished uploading cache content to blob storage!')
hashStream.end()
sha256Hash = hashStream.read() as string
core.info(`SHA256 hash of uploaded artifact zip is ${sha256Hash}`)
core.info(`Uploaded: ${uploadByteCount} bytes`)
if (uploadByteCount === 0) {
core.error(
`No data was uploaded to blob storage. Reported upload byte count is 0.`
)
}
return {
uploadSize: uploadByteCount,
sha256Hash
}
}
export async function UploadCacheFile(
uploadURL: GetCacheBlobUploadURLResponse,
archivePath: string,
): Promise<{}> {
core.info(`Uploading ${archivePath} to: ${JSON.stringify(uploadURL)}`)
// Specify data transfer options
const uploadOptions: BlockBlobParallelUploadOptions = {
blockSize: 4 * 1024 * 1024, // 4 MiB max block size
concurrency: 2, // maximum number of parallel transfer workers
maxSingleShotSize: 8 * 1024 * 1024, // 8 MiB initial transfer size
};
// const blobClient: BlobClient = new BlobClient(uploadURL.urls[0])
const blobClient: BlobClient = new BlobClient(uploadURL.urls[0].url)
const blockBlobClient: BlockBlobClient = blobClient.getBlockBlobClient()
core.info(`BlobClient: ${JSON.stringify(blobClient)}`)
core.info(`blockBlobClient: ${JSON.stringify(blockBlobClient)}`)
return blockBlobClient.uploadFile(archivePath, uploadOptions);
}

View File

@ -1,26 +0,0 @@
import * as core from '@actions/core'
import {GetCacheBlobUploadURLResponse} from '../../../generated/results/api/v1/blobcache'
import {BlobClient, BlockBlobClient, BlockBlobParallelUploadOptions} from '@azure/storage-blob'
export async function UploadCache(
uploadURL: GetCacheBlobUploadURLResponse,
archivePath: string,
): Promise<{}> {
core.info(`Uploading ${archivePath} to: ${JSON.stringify(uploadURL)}`)
// Specify data transfer options
const uploadOptions: BlockBlobParallelUploadOptions = {
blockSize: 4 * 1024 * 1024, // 4 MiB max block size
concurrency: 2, // maximum number of parallel transfer workers
maxSingleShotSize: 8 * 1024 * 1024, // 8 MiB initial transfer size
};
// const blobClient: BlobClient = new BlobClient(uploadURL.urls[0])
const blobClient: BlobClient = new BlobClient(uploadURL.urls[0].url)
const blockBlobClient: BlockBlobClient = blobClient.getBlockBlobClient()
core.info(`BlobClient: ${JSON.stringify(blobClient)}`)
core.info(`blockBlobClient: ${JSON.stringify(blockBlobClient)}`)
return blockBlobClient.uploadFile(archivePath, uploadOptions);
}

0
packages/cache/src/internal/v2/zip.ts vendored Normal file
View File