glitchier-soc/lib/mastodon/domains_cli.rb
Claire cb4e28f405
Add tootctl domains purge options to select subdomains and keep domain blocks (#22063)
* Add --include-subdomains option to tootctl domains purge

* Add support for '*.' subdomain wildcard patterns in `tootctl domains purge`

* Fix custom emojis deletion not following subdomain and URI options

* Change `tootctl domains purge` to not purge domain blocks unless --purge-domain-blocks is passed

* Refactor `tootctl domains purge`

* Add feedback on deleted domain blocks
2023-01-18 16:50:50 +01:00

223 lines
9.7 KiB
Ruby

# frozen_string_literal: true
require 'concurrent'
require_relative '../../config/boot'
require_relative '../../config/environment'
require_relative 'cli_helper'
module Mastodon
class DomainsCLI < Thor
include CLIHelper
def self.exit_on_failure?
true
end
option :concurrency, type: :numeric, default: 5, aliases: [:c]
option :verbose, type: :boolean, aliases: [:v]
option :dry_run, type: :boolean
option :limited_federation_mode, type: :boolean
option :by_uri, type: :boolean
option :include_subdomains, type: :boolean
option :purge_domain_blocks, type: :boolean
desc 'purge [DOMAIN...]', 'Remove accounts from a DOMAIN without a trace'
long_desc <<-LONG_DESC
Remove all accounts from a given DOMAIN without leaving behind any
records. Unlike a suspension, if the DOMAIN still exists in the wild,
it means the accounts could return if they are resolved again.
When the --limited-federation-mode option is given, instead of purging accounts
from a single domain, all accounts from domains that have not been explicitly allowed
are removed from the database.
When the --by-uri option is given, DOMAIN is used to match the domain part of actor
URIs rather than the domain part of the webfinger handle. For instance, an account
that has the handle `foo@bar.com` but whose profile is at the URL
`https://mastodon-bar.com/users/foo`, would be purged by either
`tootctl domains purge bar.com` or `tootctl domains purge --by-uri mastodon-bar.com`.
When the --include-subdomains option is given, not only DOMAIN is deleted, but all
subdomains as well. Note that this may be considerably slower.
When the --purge-domain-blocks option is given, also purge matching domain blocks.
LONG_DESC
def purge(*domains)
dry_run = options[:dry_run] ? ' (DRY RUN)' : ''
domains = domains.map { |domain| TagManager.instance.normalize_domain(domain) }
account_scope = Account.none
domain_block_scope = DomainBlock.none
emoji_scope = CustomEmoji.none
# Sanity check on command arguments
if options[:limited_federation_mode] && !domains.empty?
say('DOMAIN parameter not supported with --limited-federation-mode', :red)
exit(1)
elsif domains.empty? && !options[:limited_federation_mode]
say('No domain(s) given', :red)
exit(1)
end
# Build scopes from command arguments
if options[:limited_federation_mode]
account_scope = Account.remote.where.not(domain: DomainAllow.select(:domain))
emoji_scope = CustomEmoji.remote.where.not(domain: DomainAllow.select(:domain))
else
# Handle wildcard subdomains
subdomain_patterns = domains.filter_map { |domain| "%.#{Account.sanitize_sql_like(domain[2..])}" if domain.start_with?('*.') }
domains = domains.filter { |domain| !domain.start_with?('*.') }
# Handle --include-subdomains
subdomain_patterns += domains.map { |domain| "%.#{Account.sanitize_sql_like(domain)}" } if options[:include_subdomains]
uri_patterns = (domains.map { |domain| Account.sanitize_sql_like(domain) } + subdomain_patterns).map { |pattern| "https://#{pattern}/%" }
if options[:purge_domain_blocks]
domain_block_scope = DomainBlock.where(domain: domains)
domain_block_scope = domain_block_scope.or(DomainBlock.where(DomainBlock.arel_table[:domain].matches_any(subdomain_patterns))) unless subdomain_patterns.empty?
end
if options[:by_uri]
account_scope = Account.remote.where(Account.arel_table[:uri].matches_any(uri_patterns, false, true))
emoji_scope = CustomEmoji.remote.where(CustomEmoji.arel_table[:uri].matches_any(uri_patterns, false, true))
else
account_scope = Account.remote.where(domain: domains)
account_scope = account_scope.or(Account.remote.where(Account.arel_table[:domain].matches_any(subdomain_patterns))) unless subdomain_patterns.empty?
emoji_scope = CustomEmoji.where(domain: domains)
emoji_scope = emoji_scope.or(CustomEmoji.remote.where(CustomEmoji.arel_table[:uri].matches_any(subdomain_patterns))) unless subdomain_patterns.empty?
end
end
# Actually perform the deletions
processed, = parallelize_with_progress(account_scope) do |account|
DeleteAccountService.new.call(account, reserve_username: false, skip_side_effects: true) unless options[:dry_run]
end
say("Removed #{processed} accounts#{dry_run}", :green)
if options[:purge_domain_blocks]
domain_block_count = domain_block_scope.count
domain_block_scope.in_batches.destroy_all unless options[:dry_run]
say("Removed #{domain_block_count} domain blocks#{dry_run}", :green)
end
custom_emojis_count = emoji_scope.count
emoji_scope.in_batches.destroy_all unless options[:dry_run]
Instance.refresh unless options[:dry_run]
say("Removed #{custom_emojis_count} custom emojis#{dry_run}", :green)
end
option :concurrency, type: :numeric, default: 50, aliases: [:c]
option :format, type: :string, default: 'summary', aliases: [:f]
option :exclude_suspended, type: :boolean, default: false, aliases: [:x]
desc 'crawl [START]', 'Crawl all known peers, optionally beginning at START'
long_desc <<-LONG_DESC
Crawl the fediverse by using the Mastodon REST API endpoints that expose
all known peers, and collect statistics from those peers, as long as those
peers support those API endpoints. When no START is given, the command uses
this server's own database of known peers to seed the crawl.
The --concurrency (-c) option controls the number of threads performing HTTP
requests at the same time. More threads means the crawl may complete faster.
The --format (-f) option controls how the data is displayed at the end. By
default (`summary`), a summary of the statistics is returned. The other options
are `domains`, which returns a newline-delimited list of all discovered peers,
and `json`, which dumps all the aggregated data raw.
The --exclude-suspended (-x) option means that domains that are suspended
instance-wide do not appear in the output and are not included in summaries.
This also excludes subdomains of any of those domains.
LONG_DESC
def crawl(start = nil)
stats = Concurrent::Hash.new
processed = Concurrent::AtomicFixnum.new(0)
failed = Concurrent::AtomicFixnum.new(0)
start_at = Time.now.to_f
seed = start ? [start] : Instance.pluck(:domain)
blocked_domains = /\.?(#{DomainBlock.where(severity: 1).pluck(:domain).map { |domain| Regexp.escape(domain) }.join('|')})$/
progress = create_progress_bar
pool = Concurrent::ThreadPoolExecutor.new(min_threads: 0, max_threads: options[:concurrency], idletime: 10, auto_terminate: true, max_queue: 0)
work_unit = ->(domain) do
next if stats.key?(domain)
next if options[:exclude_suspended] && domain.match?(blocked_domains)
stats[domain] = nil
begin
Request.new(:get, "https://#{domain}/api/v1/instance").perform do |res|
next unless res.code == 200
stats[domain] = Oj.load(res.to_s)
end
Request.new(:get, "https://#{domain}/api/v1/instance/peers").perform do |res|
next unless res.code == 200
Oj.load(res.to_s).reject { |peer| stats.key?(peer) }.each do |peer|
pool.post(peer, &work_unit)
end
end
Request.new(:get, "https://#{domain}/api/v1/instance/activity").perform do |res|
next unless res.code == 200
stats[domain]['activity'] = Oj.load(res.to_s)
end
rescue StandardError
failed.increment
ensure
processed.increment
progress.increment unless progress.finished?
end
end
seed.each do |domain|
pool.post(domain, &work_unit)
end
sleep 20
sleep 20 until pool.queue_length.zero?
pool.shutdown
pool.wait_for_termination(20)
ensure
progress.finish
pool.shutdown
case options[:format]
when 'summary'
stats_to_summary(stats, processed, failed, start_at)
when 'domains'
stats_to_domains(stats)
when 'json'
stats_to_json(stats)
end
end
private
def stats_to_summary(stats, processed, failed, start_at)
stats.compact!
total_domains = stats.size
total_users = stats.reduce(0) { |sum, (_key, val)| val.is_a?(Hash) && val['stats'].is_a?(Hash) ? sum + val['stats']['user_count'].to_i : sum }
total_active = stats.reduce(0) { |sum, (_key, val)| val.is_a?(Hash) && val['activity'].is_a?(Array) && val['activity'].size > 2 && val['activity'][1].is_a?(Hash) ? sum + val['activity'][1]['logins'].to_i : sum }
total_joined = stats.reduce(0) { |sum, (_key, val)| val.is_a?(Hash) && val['activity'].is_a?(Array) && val['activity'].size > 2 && val['activity'][1].is_a?(Hash) ? sum + val['activity'][1]['registrations'].to_i : sum }
say("Visited #{processed.value} domains, #{failed.value} failed (#{(Time.now.to_f - start_at).round}s elapsed)", :green)
say("Total servers: #{total_domains}", :green)
say("Total registered: #{total_users}", :green)
say("Total active last week: #{total_active}", :green)
say("Total joined last week: #{total_joined}", :green)
end
def stats_to_domains(stats)
say(stats.keys.join("\n"))
end
def stats_to_json(stats)
stats.compact!
say(Oj.dump(stats))
end
end
end