overleaf/services/docstore/app/js/DocArchiveManager.js

257 lines
6.7 KiB
JavaScript
Raw Normal View History

2020-07-31 11:20:07 -04:00
const { callbackify } = require('util')
const MongoManager = require('./MongoManager').promises
const Errors = require('./Errors')
const logger = require('@overleaf/logger')
const Settings = require('@overleaf/settings')
const crypto = require('crypto')
const { ReadableString } = require('@overleaf/stream-utils')
const RangeManager = require('./RangeManager')
2020-07-23 14:42:49 -04:00
const PersistorManager = require('./PersistorManager')
2020-09-14 11:34:34 -04:00
const pMap = require('p-map')
2015-06-02 14:55:22 -04:00
const Bson = require('bson')
const BSON = new Bson()
const PARALLEL_JOBS = Settings.parallelArchiveJobs
const UN_ARCHIVE_BATCH_SIZE = Settings.unArchiveBatchSize
2015-06-02 14:55:22 -04:00
2020-07-23 14:42:49 -04:00
module.exports = {
archiveAllDocs: callbackify(archiveAllDocs),
archiveDoc: callbackify(archiveDoc),
unArchiveAllDocs: callbackify(unArchiveAllDocs),
unarchiveDoc: callbackify(unarchiveDoc),
destroyProject: callbackify(destroyProject),
getDoc: callbackify(getDoc),
2020-07-23 14:42:49 -04:00
promises: {
archiveAllDocs,
archiveDoc,
unArchiveAllDocs,
unarchiveDoc,
destroyProject,
getDoc,
2021-07-13 07:04:48 -04:00
},
2020-07-23 14:42:49 -04:00
}
2015-06-02 14:55:22 -04:00
2020-07-23 14:42:49 -04:00
async function archiveAllDocs(projectId) {
if (!_isArchivingEnabled()) {
return
}
const docIds = await MongoManager.getNonArchivedProjectDocIds(projectId)
await pMap(docIds, docId => archiveDoc(projectId, docId), {
concurrency: PARALLEL_JOBS,
})
2020-07-23 14:42:49 -04:00
}
2015-06-02 14:55:22 -04:00
async function archiveDoc(projectId, docId) {
if (!_isArchivingEnabled()) {
return
}
const doc = await MongoManager.getDocForArchiving(projectId, docId)
if (!doc) {
// The doc wasn't found, it was already archived, or the lock couldn't be
// acquired. Since we don't know which it is, silently return.
return
}
logger.debug({ projectId, docId: doc._id }, 'sending doc to persistor')
2020-07-23 14:42:49 -04:00
const key = `${projectId}/${doc._id}`
2020-07-23 14:42:49 -04:00
if (doc.lines == null) {
throw new Error('doc has no lines')
}
// warn about any oversized docs already in mongo
const linesSize = BSON.calculateObjectSize(doc.lines || {})
const rangesSize = BSON.calculateObjectSize(doc.ranges || {})
if (
linesSize > Settings.max_doc_length ||
rangesSize > Settings.max_doc_length
) {
logger.warn(
{ projectId, docId: doc._id, linesSize, rangesSize },
'large doc found when archiving project'
)
}
2020-07-23 14:42:49 -04:00
const json = JSON.stringify({
lines: doc.lines,
ranges: doc.ranges,
rev: doc.rev,
2021-07-13 07:04:48 -04:00
schema_v: 1,
2020-07-23 14:42:49 -04:00
})
2020-07-23 14:42:49 -04:00
// this should never happen, but protects against memory-corruption errors that
// have happened in the past
if (json.indexOf('\u0000') > -1) {
const error = new Error('null bytes detected')
logger.err({ err: error, doc }, error.message)
throw error
}
const md5 = crypto.createHash('md5').update(json).digest('hex')
const stream = new ReadableString(json)
await PersistorManager.sendStream(Settings.docstore.bucket, key, stream, {
2021-07-13 07:04:48 -04:00
sourceMd5: md5,
2020-07-23 14:42:49 -04:00
})
await MongoManager.markDocAsArchived(projectId, docId, doc.rev)
2020-07-23 14:42:49 -04:00
}
async function unArchiveAllDocs(projectId) {
if (!_isArchivingEnabled()) {
return
}
while (true) {
let docs
if (Settings.docstore.keepSoftDeletedDocsArchived) {
docs = await MongoManager.getNonDeletedArchivedProjectDocs(
projectId,
UN_ARCHIVE_BATCH_SIZE
)
} else {
docs = await MongoManager.getArchivedProjectDocs(
projectId,
UN_ARCHIVE_BATCH_SIZE
)
}
if (!docs || docs.length === 0) {
break
}
2021-07-13 07:04:48 -04:00
await pMap(docs, doc => unarchiveDoc(projectId, doc._id), {
concurrency: PARALLEL_JOBS,
})
2020-07-23 14:42:49 -04:00
}
}
// get the doc from the PersistorManager without storing it in mongo
async function getDoc(projectId, docId) {
2020-07-23 14:42:49 -04:00
const key = `${projectId}/${docId}`
const sourceMd5 = await PersistorManager.getObjectMd5Hash(
Settings.docstore.bucket,
key
)
const stream = await PersistorManager.getObjectStream(
Settings.docstore.bucket,
key
)
2020-07-23 14:42:49 -04:00
stream.resume()
const buffer = await _streamToBuffer(projectId, docId, stream)
const md5 = crypto.createHash('md5').update(buffer).digest('hex')
2020-07-23 14:42:49 -04:00
if (sourceMd5 !== md5) {
throw new Errors.Md5MismatchError('md5 mismatch when downloading doc', {
key,
sourceMd5,
2021-07-13 07:04:48 -04:00
md5,
})
2020-07-23 14:42:49 -04:00
}
return _deserializeArchivedDoc(buffer)
}
// get the doc and unarchive it to mongo
async function unarchiveDoc(projectId, docId) {
logger.debug({ projectId, docId }, 'getting doc from persistor')
const mongoDoc = await MongoManager.findDoc(projectId, docId, {
inS3: 1,
rev: 1,
})
if (!mongoDoc.inS3) {
// The doc is already unarchived
return
}
if (!_isArchivingEnabled()) {
throw new Error(
'found archived doc, but archiving backend is not configured'
)
}
const archivedDoc = await getDoc(projectId, docId)
if (archivedDoc.rev == null) {
// Older archived docs didn't have a rev. Assume that the rev of the
// archived doc is the rev that was stored in Mongo when we retrieved it
// earlier.
archivedDoc.rev = mongoDoc.rev
}
await MongoManager.restoreArchivedDoc(projectId, docId, archivedDoc)
2020-07-23 14:42:49 -04:00
}
async function destroyProject(projectId) {
const tasks = [MongoManager.destroyProject(projectId)]
if (_isArchivingEnabled()) {
tasks.push(
PersistorManager.deleteDirectory(Settings.docstore.bucket, projectId)
)
2020-07-23 14:42:49 -04:00
}
await Promise.all(tasks)
2020-07-23 14:42:49 -04:00
}
async function _streamToBuffer(projectId, docId, stream) {
const chunks = []
let size = 0
let logged = false
const logIfTooLarge = finishedReading => {
if (size <= Settings.max_doc_length) return
// Log progress once and then again at the end.
if (logged && !finishedReading) return
logger.warn(
{ projectId, docId, size, finishedReading },
'potentially large doc pulled down from gcs'
)
logged = true
}
return await new Promise((resolve, reject) => {
stream.on('data', chunk => {
size += chunk.byteLength
logIfTooLarge(false)
chunks.push(chunk)
})
stream.on('error', reject)
stream.on('end', () => {
logIfTooLarge(true)
resolve(Buffer.concat(chunks))
})
2020-07-23 14:42:49 -04:00
})
}
function _deserializeArchivedDoc(buffer) {
const doc = JSON.parse(buffer)
const result = {}
if (doc.schema_v === 1 && doc.lines != null) {
result.lines = doc.lines
if (doc.ranges != null) {
result.ranges = RangeManager.jsonRangesToMongo(doc.ranges)
}
} else if (Array.isArray(doc)) {
result.lines = doc
} else {
throw new Error("I don't understand the doc format in s3")
}
2020-07-23 14:42:49 -04:00
if (doc.rev != null) {
result.rev = doc.rev
2020-07-23 14:42:49 -04:00
}
return result
2020-07-23 14:42:49 -04:00
}
function _isArchivingEnabled() {
const backend = Settings.docstore.backend
if (!backend) {
return false
}
// The default backend is S3. If another backend is configured or the S3
// backend itself is correctly configured, then archiving is enabled.
if (backend === 's3' && Settings.docstore.s3 == null) {
return false
}
return true
}