akkoma/lib/pleroma/migrators/hashtags_table_migrator.ex

372 lines
11 KiB
Elixir

# Pleroma: A lightweight social networking server
# Copyright © 2017-2021 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Migrators.HashtagsTableMigrator do
use GenServer
require Logger
import Ecto.Query
alias __MODULE__.State
alias Pleroma.Config
alias Pleroma.Hashtag
alias Pleroma.Object
alias Pleroma.Repo
defdelegate data_migration(), to: Pleroma.DataMigration, as: :populate_hashtags_table
defdelegate data_migration_id(), to: State
defdelegate state(), to: State
defdelegate persist_state(), to: State, as: :persist_to_db
defdelegate get_stat(key, value \\ nil), to: State, as: :get_data_key
defdelegate put_stat(key, value), to: State, as: :put_data_key
defdelegate increment_stat(key, increment), to: State, as: :increment_data_key
@feature_config_path [:database, :improved_hashtag_timeline]
@reg_name {:global, __MODULE__}
def whereis, do: GenServer.whereis(@reg_name)
def feature_state, do: Config.get(@feature_config_path)
def start_link(_) do
case whereis() do
nil ->
GenServer.start_link(__MODULE__, nil, name: @reg_name)
pid ->
{:ok, pid}
end
end
@impl true
def init(_) do
{:ok, nil, {:continue, :init_state}}
end
@impl true
def handle_continue(:init_state, _state) do
{:ok, _} = State.start_link(nil)
data_migration = data_migration()
manual_migrations = Config.get([:instance, :manual_data_migrations], [])
cond do
Config.get(:env) == :test ->
update_status(:noop)
is_nil(data_migration) ->
message = "Data migration does not exist."
update_status(:failed, message)
Logger.error("#{__MODULE__}: #{message}")
data_migration.state == :manual or data_migration.name in manual_migrations ->
message = "Data migration is in manual execution or manual fix mode."
update_status(:manual, message)
Logger.warn("#{__MODULE__}: #{message}")
data_migration.state == :complete ->
on_complete(data_migration)
true ->
send(self(), :migrate_hashtags)
end
{:noreply, nil}
end
@impl true
def handle_info(:migrate_hashtags, state) do
State.reinit()
update_status(:running)
put_stat(:started_at, NaiveDateTime.utc_now())
data_migration_id = data_migration_id()
max_processed_id = get_stat(:max_processed_id, 0)
Logger.info("Transferring embedded hashtags to `hashtags` (from oid: #{max_processed_id})...")
query()
|> where([object], object.id > ^max_processed_id)
|> Repo.chunk_stream(100, :batches, timeout: :infinity)
|> Stream.each(fn objects ->
object_ids = Enum.map(objects, & &1.id)
results = Enum.map(objects, &transfer_object_hashtags(&1))
failed_ids =
results
|> Enum.filter(&(elem(&1, 0) == :error))
|> Enum.map(&elem(&1, 1))
# Count of objects with hashtags (`{:noop, id}` is returned for objects having other AS2 tags)
chunk_affected_count =
results
|> Enum.filter(&(elem(&1, 0) == :ok))
|> length()
for failed_id <- failed_ids do
_ =
Repo.query(
"INSERT INTO data_migration_failed_ids(data_migration_id, record_id) " <>
"VALUES ($1, $2) ON CONFLICT DO NOTHING;",
[data_migration_id, failed_id]
)
end
_ =
Repo.query(
"DELETE FROM data_migration_failed_ids " <>
"WHERE data_migration_id = $1 AND record_id = ANY($2)",
[data_migration_id, object_ids -- failed_ids]
)
max_object_id = Enum.at(object_ids, -1)
put_stat(:max_processed_id, max_object_id)
increment_stat(:processed_count, length(object_ids))
increment_stat(:failed_count, length(failed_ids))
increment_stat(:affected_count, chunk_affected_count)
put_stat(:records_per_second, records_per_second())
persist_state()
# A quick and dirty approach to controlling the load this background migration imposes
sleep_interval = Config.get([:populate_hashtags_table, :sleep_interval_ms], 0)
Process.sleep(sleep_interval)
end)
|> Stream.run()
fault_rate = fault_rate()
put_stat(:fault_rate, fault_rate)
fault_rate_allowance = Config.get([:populate_hashtags_table, :fault_rate_allowance], 0)
cond do
fault_rate == 0 ->
set_complete()
is_float(fault_rate) and fault_rate <= fault_rate_allowance ->
message = """
Done with fault rate of #{fault_rate} which doesn't exceed #{fault_rate_allowance}.
Putting data migration to manual fix mode. Check `retry_failed/0`.
"""
Logger.warn("#{__MODULE__}: #{message}")
update_status(:manual, message)
on_complete(data_migration())
true ->
message = "Too many failures. Check data_migration_failed_ids records / `retry_failed/0`."
Logger.error("#{__MODULE__}: #{message}")
update_status(:failed, message)
end
persist_state()
{:noreply, state}
end
def fault_rate do
with failures_count when is_integer(failures_count) <- failures_count() do
failures_count / Enum.max([get_stat(:affected_count, 0), 1])
else
_ -> :error
end
end
defp records_per_second do
get_stat(:processed_count, 0) / Enum.max([running_time(), 1])
end
defp running_time do
NaiveDateTime.diff(NaiveDateTime.utc_now(), get_stat(:started_at, NaiveDateTime.utc_now()))
end
@hashtags_objects_cleanup_query """
DELETE FROM hashtags_objects WHERE object_id IN
(SELECT DISTINCT objects.id FROM objects
JOIN hashtags_objects ON hashtags_objects.object_id = objects.id LEFT JOIN activities
ON COALESCE(activities.data->'object'->>'id', activities.data->>'object') =
(objects.data->>'id')
AND activities.data->>'type' = 'Create'
WHERE activities.id IS NULL);
"""
@hashtags_cleanup_query """
DELETE FROM hashtags WHERE id IN
(SELECT hashtags.id FROM hashtags
LEFT OUTER JOIN hashtags_objects
ON hashtags_objects.hashtag_id = hashtags.id
WHERE hashtags_objects.hashtag_id IS NULL);
"""
@doc """
Deletes `hashtags_objects` for legacy objects not asoociated with Create activity.
Also deletes unreferenced `hashtags` records (might occur after deletion of `hashtags_objects`).
"""
def delete_non_create_activities_hashtags do
{:ok, %{num_rows: hashtags_objects_count}} =
Repo.query(@hashtags_objects_cleanup_query, [], timeout: :infinity)
{:ok, %{num_rows: hashtags_count}} =
Repo.query(@hashtags_cleanup_query, [], timeout: :infinity)
{:ok, hashtags_objects_count, hashtags_count}
end
defp query do
# Note: most objects have Mention-type AS2 tags and no hashtags (but we can't filter them out)
# Note: not checking activity type, expecting remove_non_create_objects_hashtags/_ to clean up
from(
object in Object,
where:
fragment("(?)->'tag' IS NOT NULL AND (?)->'tag' != '[]'::jsonb", object.data, object.data),
select: %{
id: object.id,
tag: fragment("(?)->'tag'", object.data)
}
)
|> join(:left, [o], hashtags_objects in fragment("SELECT object_id FROM hashtags_objects"),
on: hashtags_objects.object_id == o.id
)
|> where([_o, hashtags_objects], is_nil(hashtags_objects.object_id))
end
@spec transfer_object_hashtags(Map.t()) :: {:noop | :ok | :error, integer()}
defp transfer_object_hashtags(object) do
embedded_tags = if Map.has_key?(object, :tag), do: object.tag, else: object.data["tag"]
hashtags = Object.object_data_hashtags(%{"tag" => embedded_tags})
if Enum.any?(hashtags) do
transfer_object_hashtags(object, hashtags)
else
{:noop, object.id}
end
end
defp transfer_object_hashtags(object, hashtags) do
Repo.transaction(fn ->
with {:ok, hashtag_records} <- Hashtag.get_or_create_by_names(hashtags) do
maps = Enum.map(hashtag_records, &%{hashtag_id: &1.id, object_id: object.id})
base_error = "ERROR when inserting hashtags_objects for object with id #{object.id}"
try do
with {rows_count, _} when is_integer(rows_count) <-
Repo.insert_all("hashtags_objects", maps, on_conflict: :nothing) do
object.id
else
e ->
Logger.error("#{base_error}: #{inspect(e)}")
Repo.rollback(object.id)
end
rescue
e ->
Logger.error("#{base_error}: #{inspect(e)}")
Repo.rollback(object.id)
end
else
e ->
error = "ERROR: could not create hashtags for object #{object.id}: #{inspect(e)}"
Logger.error(error)
Repo.rollback(object.id)
end
end)
end
@doc "Approximate count for current iteration (including processed records count)"
def count(force \\ false, timeout \\ :infinity) do
stored_count = get_stat(:count)
if stored_count && !force do
stored_count
else
processed_count = get_stat(:processed_count, 0)
max_processed_id = get_stat(:max_processed_id, 0)
query = where(query(), [object], object.id > ^max_processed_id)
count = Repo.aggregate(query, :count, :id, timeout: timeout) + processed_count
put_stat(:count, count)
persist_state()
count
end
end
defp on_complete(data_migration) do
cond do
data_migration.feature_lock ->
:noop
not is_nil(feature_state()) ->
:noop
true ->
Config.put(@feature_config_path, true)
:ok
end
end
def failed_objects_query do
from(o in Object)
|> join(:inner, [o], dmf in fragment("SELECT * FROM data_migration_failed_ids"),
on: dmf.record_id == o.id
)
|> where([_o, dmf], dmf.data_migration_id == ^data_migration_id())
|> order_by([o], asc: o.id)
end
def failures_count do
with {:ok, %{rows: [[count]]}} <-
Repo.query(
"SELECT COUNT(record_id) FROM data_migration_failed_ids WHERE data_migration_id = $1;",
[data_migration_id()]
) do
count
end
end
def retry_failed do
data_migration_id = data_migration_id()
failed_objects_query()
|> Repo.chunk_stream(100, :one)
|> Stream.each(fn object ->
with {res, _} when res != :error <- transfer_object_hashtags(object) do
_ =
Repo.query(
"DELETE FROM data_migration_failed_ids " <>
"WHERE data_migration_id = $1 AND record_id = $2",
[data_migration_id, object.id]
)
end
end)
|> Stream.run()
put_stat(:failed_count, failures_count())
persist_state()
force_continue()
end
def force_continue do
send(whereis(), :migrate_hashtags)
end
def force_restart do
:ok = State.reset()
force_continue()
end
def set_complete do
update_status(:complete)
persist_state()
on_complete(data_migration())
end
defp update_status(status, message \\ nil) do
put_stat(:state, status)
put_stat(:message, message)
end
end