Fix temp files generated during the upload process not being cleaned up quickly enough. This included downloaded files, generated preview images, and Ugoira video conversions. Before we relied on `Tempfile` cleaning up files automatically. But this only happened when the Tempfile object was garbage collected, which could take a long time. In the meantime we could have hundreds of megabytes of temp files hanging around. The fix is to explicitly close temp files when we're done with them. But the standard `Tempfile` class doesn't immediately delete the file when it's closed. So we also have to introduce a Danbooru::Tempfile wrapper that deletes the tempfile as soon as it's closed.
120 lines
4.0 KiB
Ruby
120 lines
4.0 KiB
Ruby
# frozen_string_literal: true
|
|
|
|
# Perform a daily database dump to BigQuery and to Google Cloud Storage. This
|
|
# contains all data visible to anonymous users.
|
|
#
|
|
# The database dumps are publicly accessible. The BigQuery data is at
|
|
# `danbooru1.danbooru_public.{table}`. The Google Cloud Storage data is at
|
|
# `gs://danbooru_public/data/{table}.json`. The storage bucket contains the data
|
|
# in newline-delimited JSON format.
|
|
#
|
|
# @see DanbooruMaintenance#daily
|
|
# @see https://console.cloud.google.com/storage/browser/danbooru_public
|
|
# @see https://console.cloud.google.com/bigquery?d=danbooru_public&p=danbooru1&t=posts&page=table
|
|
# @see https://cloud.google.com/bigquery/docs
|
|
# @see https://cloud.google.com/storage/docs
|
|
# @see https://en.wikipedia.org/wiki/JSON_streaming#Line-delimited_JSON
|
|
class BigqueryExportService
|
|
extend Memoist
|
|
|
|
attr_reader :model, :dataset_name, :credentials
|
|
|
|
# Prepare to dump a table. Call {#export!} to dump it.
|
|
# @param model [ApplicationRecord] the database table to dump
|
|
# @param dataset_name [String] the BigQuery dataset name
|
|
# @param credentials [String] the Google Cloud credentials (in JSON format)
|
|
def initialize(model = nil, dataset_name: "danbooru_public", credentials: default_credentials)
|
|
@model = model
|
|
@dataset_name = dataset_name
|
|
@credentials = credentials
|
|
end
|
|
|
|
# Start a background job for each table to export it to BigQuery.
|
|
def self.async_export_all!(**options)
|
|
models.each do |model|
|
|
BigqueryExportJob.perform_later(model: model, **options)
|
|
end
|
|
end
|
|
|
|
# The list of database tables to dump.
|
|
def self.models
|
|
Rails.application.eager_load!
|
|
|
|
models = ApplicationRecord.descendants.sort_by(&:name)
|
|
models -= [GoodJob::BaseRecord, GoodJob::Process, GoodJob::Execution, GoodJob::ActiveJobJob, GoodJob::Job, GoodJob::Setting, TagRelationship, ArtistVersion, ArtistCommentaryVersion, NoteVersion, PoolVersion, PostVersion, WikiPageVersion, Post, PostEvent, PostVote, MediaAsset, Favorite, AITag, UserAction]
|
|
models
|
|
end
|
|
|
|
def enabled?
|
|
credentials.present?
|
|
end
|
|
|
|
# Dump the table to Cloud Storage and BigQuery.
|
|
def export!
|
|
return unless enabled? && records.any?
|
|
|
|
file = dump_records!
|
|
upload_to_bigquery!(file)
|
|
ensure
|
|
file&.close
|
|
end
|
|
|
|
# Dump the table's records to a gzipped, newline-delimited JSON tempfile.
|
|
def dump_records!(file = Danbooru::Tempfile.new("danbooru-export-dump-#{model.name}-", binmode: true))
|
|
gzip = Zlib::GzipWriter.new(file)
|
|
|
|
CurrentUser.scoped(User.anonymous) do
|
|
records.find_each(batch_size: 5_000) do |record|
|
|
gzip.puts(record.to_json)
|
|
end
|
|
end
|
|
|
|
gzip.finish
|
|
file
|
|
end
|
|
|
|
# Upload the JSON dump to Cloud Storage, then load it into BigQuery.
|
|
def upload_to_bigquery!(file)
|
|
table_name = model.model_name.collection
|
|
gsfilename = "data/#{table_name}.json"
|
|
|
|
gsfile = bucket.create_file(file.path, gsfilename, content_encoding: "gzip")
|
|
job = dataset.load_job(table_name, gsfile, format: "json", autodetect: true, create: "needed", write: "truncate")
|
|
|
|
job.wait_until_done!
|
|
job
|
|
end
|
|
|
|
# The list of records to dump.
|
|
def records
|
|
model.visible(User.anonymous)
|
|
end
|
|
|
|
# Find or create the BigQuery dataset.
|
|
def dataset
|
|
bigquery.dataset(dataset_name) || bigquery.create_dataset(dataset_name)
|
|
end
|
|
|
|
# Find or create the Google Storage bucket.
|
|
def bucket
|
|
storage.bucket(dataset_name, user_project: true) || storage.create_bucket(dataset_name, acl: "public", default_acl: "public", storage_class: "standard", location: "us-east1", requester_pays: true, user_project: true)
|
|
end
|
|
|
|
# The BigQuery API client.
|
|
def bigquery
|
|
Google::Cloud::Bigquery.new(credentials: credentials)
|
|
end
|
|
|
|
# The Cloud Storage API client.
|
|
def storage
|
|
Google::Cloud::Storage.new(credentials: credentials)
|
|
end
|
|
|
|
def default_credentials
|
|
return nil unless Danbooru.config.google_cloud_credentials.present?
|
|
JSON.parse(Danbooru.config.google_cloud_credentials)
|
|
end
|
|
|
|
memoize :dataset, :bucket, :bigquery, :storage, :default_credentials
|
|
end
|