From 3e9d794ea50986db5647b4e05a408bf0208bbfa1 Mon Sep 17 00:00:00 2001 From: Eugen Rochko Date: Sat, 25 Feb 2017 03:34:37 +0100 Subject: [PATCH] Add tuning documentation, add tags back to most salmons, make status pagination headers generation more lax about next page existing --- app/controllers/api/v1/accounts_controller.rb | 4 +- .../api/v1/notifications_controller.rb | 2 +- .../api/v1/timelines_controller.rb | 6 +- app/services/favourite_service.rb | 5 +- app/services/follow_service.rb | 10 +- app/services/unfavourite_service.rb | 5 +- app/services/unfollow_service.rb | 5 +- docs/README.md | 1 + docs/Running-Mastodon/Tuning.md | 104 ++++++++++++++++++ 9 files changed, 131 insertions(+), 11 deletions(-) create mode 100644 docs/Running-Mastodon/Tuning.md diff --git a/app/controllers/api/v1/accounts_controller.rb b/app/controllers/api/v1/accounts_controller.rb index 0d02294eb3..94dba1d038 100644 --- a/app/controllers/api/v1/accounts_controller.rb +++ b/app/controllers/api/v1/accounts_controller.rb @@ -52,7 +52,7 @@ class Api::V1::AccountsController < ApiController set_maps(@statuses) set_counters_maps(@statuses) - next_path = statuses_api_v1_account_url(max_id: @statuses.last.id) if @statuses.size == limit_param(DEFAULT_STATUSES_LIMIT) + next_path = statuses_api_v1_account_url(max_id: @statuses.last.id) unless @statuses.empty? prev_path = statuses_api_v1_account_url(since_id: @statuses.first.id) unless @statuses.empty? set_pagination_headers(next_path, prev_path) @@ -66,7 +66,7 @@ class Api::V1::AccountsController < ApiController set_maps(@statuses) set_counters_maps(@statuses) - next_path = media_statuses_api_v1_account_url(max_id: @statuses.last.id) if @statuses.size == limit_param(DEFAULT_STATUSES_LIMIT) + next_path = media_statuses_api_v1_account_url(max_id: @statuses.last.id) unless @statuses.empty? prev_path = media_statuses_api_v1_account_url(since_id: @statuses.first.id) unless @statuses.empty? set_pagination_headers(next_path, prev_path) diff --git a/app/controllers/api/v1/notifications_controller.rb b/app/controllers/api/v1/notifications_controller.rb index 877356a75e..544ba24421 100644 --- a/app/controllers/api/v1/notifications_controller.rb +++ b/app/controllers/api/v1/notifications_controller.rb @@ -17,7 +17,7 @@ class Api::V1::NotificationsController < ApiController set_counters_maps(statuses) set_account_counters_maps(@notifications.map(&:from_account)) - next_path = api_v1_notifications_url(max_id: @notifications.last.id) if @notifications.size == limit_param(DEFAULT_NOTIFICATIONS_LIMIT) + next_path = api_v1_notifications_url(max_id: @notifications.last.id) unless @notifications.empty? prev_path = api_v1_notifications_url(since_id: @notifications.first.id) unless @notifications.empty? set_pagination_headers(next_path, prev_path) diff --git a/app/controllers/api/v1/timelines_controller.rb b/app/controllers/api/v1/timelines_controller.rb index a8cc2b2880..af6e5b7df2 100644 --- a/app/controllers/api/v1/timelines_controller.rb +++ b/app/controllers/api/v1/timelines_controller.rb @@ -14,7 +14,7 @@ class Api::V1::TimelinesController < ApiController set_counters_maps(@statuses) set_account_counters_maps(@statuses.flat_map { |s| [s.account, s.reblog? ? s.reblog.account : nil] }.compact.uniq) - next_path = api_v1_home_timeline_url(max_id: @statuses.last.id) if @statuses.size == limit_param(DEFAULT_STATUSES_LIMIT) + next_path = api_v1_home_timeline_url(max_id: @statuses.last.id) unless @statuses.empty? prev_path = api_v1_home_timeline_url(since_id: @statuses.first.id) unless @statuses.empty? set_pagination_headers(next_path, prev_path) @@ -30,7 +30,7 @@ class Api::V1::TimelinesController < ApiController set_counters_maps(@statuses) set_account_counters_maps(@statuses.flat_map { |s| [s.account, s.reblog? ? s.reblog.account : nil] }.compact.uniq) - next_path = api_v1_public_timeline_url(max_id: @statuses.last.id) if @statuses.size == limit_param(DEFAULT_STATUSES_LIMIT) + next_path = api_v1_public_timeline_url(max_id: @statuses.last.id) unless @statuses.empty? prev_path = api_v1_public_timeline_url(since_id: @statuses.first.id) unless @statuses.empty? set_pagination_headers(next_path, prev_path) @@ -47,7 +47,7 @@ class Api::V1::TimelinesController < ApiController set_counters_maps(@statuses) set_account_counters_maps(@statuses.flat_map { |s| [s.account, s.reblog? ? s.reblog.account : nil] }.compact.uniq) - next_path = api_v1_hashtag_timeline_url(params[:id], max_id: @statuses.last.id) if @statuses.size == limit_param(DEFAULT_STATUSES_LIMIT) + next_path = api_v1_hashtag_timeline_url(params[:id], max_id: @statuses.last.id) unless @statuses.empty? prev_path = api_v1_hashtag_timeline_url(params[:id], since_id: @statuses.first.id) unless @statuses.empty? set_pagination_headers(next_path, prev_path) diff --git a/app/services/favourite_service.rb b/app/services/favourite_service.rb index 824729ed6d..8188983028 100644 --- a/app/services/favourite_service.rb +++ b/app/services/favourite_service.rb @@ -22,10 +22,13 @@ class FavouriteService < BaseService private def build_xml(favourite) + description = "#{favourite.account.acct} favourited a status by #{favourite.status.account.acct}" + Nokogiri::XML::Builder.new do |xml| entry(xml, true) do unique_id xml, favourite.created_at, favourite.id, 'Favourite' - title xml, "#{favourite.account.acct} favourited a status by #{favourite.status.account.acct}" + title xml, description + content xml, description author(xml) do include_author xml, favourite.account diff --git a/app/services/follow_service.rb b/app/services/follow_service.rb index d67b1bf2d6..915f95b4c4 100644 --- a/app/services/follow_service.rb +++ b/app/services/follow_service.rb @@ -55,10 +55,13 @@ class FollowService < BaseService end def build_follow_request_xml(follow_request) + description = "#{follow_request.account.acct} requested to follow #{follow_request.target_account.acct}" + Nokogiri::XML::Builder.new do |xml| entry(xml, true) do unique_id xml, follow_request.created_at, follow_request.id, 'FollowRequest' - title xml, "#{follow_request.account.acct} requested to follow #{follow_request.target_account.acct}" + title xml, description + content xml, description author(xml) do include_author xml, follow_request.account @@ -75,10 +78,13 @@ class FollowService < BaseService end def build_follow_xml(follow) + description = "#{follow.account.acct} started following #{follow.target_account.acct}" + Nokogiri::XML::Builder.new do |xml| entry(xml, true) do unique_id xml, follow.created_at, follow.id, 'Follow' - title xml, "#{follow.account.acct} started following #{follow.target_account.acct}" + title xml, description + content xml, description author(xml) do include_author xml, follow.account diff --git a/app/services/unfavourite_service.rb b/app/services/unfavourite_service.rb index 1d3e6f06dd..5f0ba42540 100644 --- a/app/services/unfavourite_service.rb +++ b/app/services/unfavourite_service.rb @@ -13,10 +13,13 @@ class UnfavouriteService < BaseService private def build_xml(favourite) + description = "#{favourite.account.acct} no longer favourites a status by #{favourite.status.account.acct}" + Nokogiri::XML::Builder.new do |xml| entry(xml, true) do unique_id xml, Time.now.utc, favourite.id, 'Favourite' - title xml, "#{favourite.account.acct} no longer favourites a status by #{favourite.status.account.acct}" + title xml, description + content xml, description author(xml) do include_author xml, favourite.account diff --git a/app/services/unfollow_service.rb b/app/services/unfollow_service.rb index 07f9b93ddc..3440da364a 100644 --- a/app/services/unfollow_service.rb +++ b/app/services/unfollow_service.rb @@ -13,10 +13,13 @@ class UnfollowService < BaseService private def build_xml(follow) + description = "#{follow.account.acct} is no longer following #{follow.target_account.acct}" + Nokogiri::XML::Builder.new do |xml| entry(xml, true) do unique_id xml, Time.now.utc, follow.id, 'Follow' - title xml, "#{follow.account.acct} is no longer following #{follow.target_account.acct}" + title xml, description + content xml, description author(xml) do include_author xml, follow.account diff --git a/docs/README.md b/docs/README.md index 5036ea22c1..d35dece148 100644 --- a/docs/README.md +++ b/docs/README.md @@ -22,6 +22,7 @@ Index - [Development guide](Running-Mastodon/Development-guide.md) - [Alternative: Development with Vagrant](Running-Mastodon/Vagrant-guide.md) - [Administration guide](Running-Mastodon/Administration-guide.md) +- [Tuning Mastodon](Running-Mastodon/Tuning.md) ### Contributing to Mastodon - [Sponsors](Contributing-to-Mastodon/Sponsors.md) diff --git a/docs/Running-Mastodon/Tuning.md b/docs/Running-Mastodon/Tuning.md new file mode 100644 index 0000000000..c4acb9927c --- /dev/null +++ b/docs/Running-Mastodon/Tuning.md @@ -0,0 +1,104 @@ +Tuning Mastodon +=============== + +Mastodon has three types of processes: + +- web +- streaming API +- background processing + +By default, the web type spawns two worker processes with 5 threads each, the streaming API is a single thread/process with 10 database pool connections, and background processing spawns one process with 5 threads. + +### Web + +The web process serves short-lived HTTP requests for most of the application. The following environment variables control it: + +- `WEB_CONCURRENCY` controls the number of worker processes +- `MAX_THREADS` controls the number of threads per process + +The default is 2 workers with 5 threads each. Threads share the memory of their parent process. Different processes allocate their own memory each. Threads in Ruby are not native threads, so it's more or less: threads equal concurrency, processes equal parallelism. A larger number of threads maxes out your CPU first, a larger number of processes maxes out your RAM first. + +These values affect how many HTTP requests can be served at the same time. When not enough threads are available, requests are queued until they can be answered. + +For a single-user instance, 1 process with 5 threads should be more than enough. + +### Streaming API + +The streaming API handles long-lived HTTP and WebSockets connections, through which clients receive real-time updates. It is a single-threaded process. By default it has a database connection pool of 10, which means 10 different database queries can run *at the same time*. The database is not heavily used in the streaming API, only for initial authentication of the request, and for some special receiver-specific filter queries when receiving new messages. At the time of writing this value cannot be reconfigured, but mostly doesn't need to. + +If you need to scale up the streaming API, spawn more separate processes on different ports (e.g. 4000, 4001, 4003, etc) and load-balance between them with nginx. + +### Background processing + +Many tasks in Mastodon are delegated to background processing to ensure the HTTP requests are fast, and to prevent HTTP request aborts from affecting the execution of those tasks. Sidekiq is a single process, with a configurable numbero of threads. By default, it is 5. That means, 5 different jobs can be executed at the same time. Others will be queued until they can be processed. + +While the amount of threads in the web process affects the responsiveness of the Mastodon instance to the end-user, the amount of threads allocated to background processing affects how quickly posts can be delivered from the author to anyone else, how soon e-mails are sent out, etc. + +The amount of threads is not controlled by an environment variable in this case, but a command line argument in the invocation of Sidekiq: + + bundle exec sidekiq -c 15 -q default -q mailers -q push + +Would start the sidekiq process with 15 threads. Please mind that each threads needs to be able to connect to the database, which means that the database pool needs to be large enough to support all the threads. The database pool size is controlled with the `DB_POOL` environment variable, and defaults to the value of `MAX_THREADS` (therefore, is 5 by default). + +You might notice that the above command specifies three queues to be processed: + +- "default" contains most tasks such as delivering messages to followers and processing incoming notifications from other instances +- "mailers" contains tasks that send e-mails +- "push" contains tasks that deliver messages to other instances + +If you wish, you could start three different processes for each queue, to ensure that even when there is a lot of tasks of one type, important tasks of other types still get executed in a timely manner. + +___ + +### How to set environment variables +#### With systemd + +In the `.service` file: + +```systemd +... +Environment="WEB_CONCURRENCY=1" +Environment="MAX_THREADS=5" +ExecStart="..." +... +``` + +Don't forget to `sudo systemctl daemon-reload` before restarting the services so that the changes would take effect! + +#### With docker-compose + +Edit `docker-compose.yml`: + +```yml +... + web: + restart: always + build: . + env_file: .env.production + environment: + - WEB_CONCURRENCY=1 + - MAX_THREADS=5 +... +``` + +Re-create the containers with `docker-compose up -d` for the changes to take effect. + +You can also scale the number of containers per "service" (where service is "web", "sidekiq" and "streaming"): + + docker-compose scale web=1 sidekiq=2 streaming=3 + +Realistically the `docker-compose.yml` file needs to be modified a bit further for the above to work, because by default it wants to bind the web container to host port 3000 and streaming container to host port 4000, of either of which there is only one on the host system. However, if you change: + +```yml +ports: + - "3000:3000" +``` + +to simply: + +```yml +ports: + - "3000" +``` + +for each service respectively, Docker will allocate random host ports of the services, allowing multiple containers to run alongside each other. But it will be on you to look up which host ports those are (e.g. with `docker ps`), and they will be different on each container restart.