Scaling up your server

Managing concurrency

Mastodon has three types of processes:

• Web (Puma)
• Streaming API
• Background processing (Sidekiq)

Web (Puma)

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

Threads share the memory of their parent process. Different processes allocate their own memory, though they share some memory via copy-on-write. 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.

In terms of throughput, more processes are better than more threads.

Streaming API

The streaming API handles long-lived HTTP and WebSockets connections, through which clients receive real-time updates. The following environment variables control it:

• STREAMING_CLUSTER_NUM controls the number of worker processes
• STREAMING_API_BASE_URL controls the base URL of the streaming API

One process can handle a reasonably high number of connections. The streaming API can be hosted on a different subdomain if you want to e.g. avoid the overhead of nginx proxying the connections.

Background processing (Sidekiq)

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 number of threads.

Number of threads

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, e.g.:

bundle exec sidekiq -c 15

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 must be at least the same as the number of threads.

Queues

Sidekiq uses different queues for tasks of varying importance, where importance is defined by how much it would impact the user experience of your server’s local users if the queue wasn’t working, in order of descending importance:

The default queues and their priorities are stored in config/sidekiq.yml, but can be overridden by the command-line invocation of Sidekiq, e.g.:

bundle exec sidekiq -q default

To run just the default queue.

The way Sidekiq works with queues, it first checks for tasks from the first queue, and if there are none, checks the next queue. This means, if the first queue is overfilled, the other queues will lag behind.

As a solution, it is possible to start different Sidekiq processes for the queues to ensure truly parallel execution, by e.g. creating multiple systemd services for Sidekiq with different arguments.

Make sure you only have one scheduler queue running!!

Transaction pooling with pgBouncer

Why you might need PgBouncer

If you start running out of available Postgres connections (the default is 100) then you may find PgBouncer to be a good solution. This document describes some common gotchas as well as good configuration defaults for Mastodon.

Note that you can check “PgHero” in the administration view to see how many Postgres connections are currently being used. Typically Mastodon uses as many connections as there are threads both in Puma, Sidekiq and the streaming API combined.

Installing PgBouncer

On Debian and Ubuntu:

sudo apt install pgbouncer

Configuring PgBouncer

Setting a password

First off, if your mastodon user in Postgres is set up without a password, you will need to set a password.

Here’s how you might reset the password:

psql -p 5432 -U mastodon mastodon_production -w

Then (obviously, use a different password than the word “password”):

ALTER USER mastodon WITH PASSWORD 'password';

Then \q to quit.

Configuring userlist.txt

Edit /etc/pgbouncer/userlist.txt

As long as you specify a user/password in pgbouncer.ini later, the values in userlist.txt do not have to correspond to real PostgreSQL roles. You can arbitrarily define users and passwords, but you can reuse the “real” credentials for simplicity’s sake. Add the mastodon user to the userlist.txt:

"mastodon" "md5d75bb2be2d7086c6148944261a00f605"

Here we’re using the md5 scheme, where the md5 password is just the md5sum of password + username with the string md5 prepended. For instance, to derive the hash for user mastodon with password password, you can do:

# ubuntu, debian, etc.
echo -n "passwordmastodon" | md5sum
# macOS, openBSD, etc.
md5 -s "passwordmastodon"

Then just add md5 to the beginning of that.

You’ll also want to create a pgbouncer admin user to log in to the PgBouncer admin database. So here’s a sample userlist.txt:

"mastodon" "md5d75bb2be2d7086c6148944261a00f605"
"pgbouncer" "md5a45753afaca0db833a6f7c7b2864b9d9"

In both cases the password is just password.

Configuring pgbouncer.ini

Edit /etc/pgbouncer/pgbouncer.ini

Add a line under [databases] listing the Postgres databases you want to connect to. Here we’ll just have PgBouncer use the same username/password and database name to connect to the underlying Postgres database:

[databases]
mastodon_production = host=127.0.0.1 port=5432 dbname=mastodon_production user=mastodon password=password

The listen_addr and listen_port tells PgBouncer which address/port to accept connections. The defaults are fine:

listen_addr = 127.0.0.1
listen_port = 6432

Put md5 as the auth_type (assuming you’re using the md5 format in userlist.txt):

Make sure the pgbouncer user is an admin:

This next part is very important! The default pooling mode is session-based, but for Mastodon we want transaction-based. In other words, a Postgres connection is created when a transaction is created and dropped when the transaction is done. So you’ll want to change the pool_mode from session to transaction:

Next up, max_client_conn defines how many connections PgBouncer itself will accept, and default_pool_size puts a limit on how many Postgres connections will be opened under the hood. (In PgHero the number of connections reported will correspond to default_pool_size because it has no knowledge of PgBouncer.)

The defaults are fine to start, and you can always increase them later:

max_client_conn = 100
default_pool_size = 20

Don’t forget to reload or restart pgbouncer after making your changes:

sudo systemctl reload pgbouncer

Debugging that it all works

You should be able to connect to PgBouncer just like you would with Postgres:

psql -p 6432 -U mastodon mastodon_production

And then use your password to log in.

You can also check the PgBouncer logs like so:

tail -f /var/log/postgresql/pgbouncer.log

Configuring Mastodon to talk to PgBouncer

In your .env.production file, first off make sure that this is set:

PREPARED_STATEMENTS=false

Since we’re using transaction-based pooling, we can’t use prepared statements.

Next up, configure Mastodon to use port 6432 (PgBouncer) instead of 5432 (Postgres) and you should be good to go:

DB_HOST=localhost
DB_USER=mastodon
DB_NAME=mastodon_production
DB_PASS=password
DB_PORT=6432

Administering PgBouncer

The easiest way to reboot is:

sudo systemctl restart pgbouncer

But if you’ve set up a PgBouncer admin user, you can also connect as the admin:

psql -p 6432 -U pgbouncer pgbouncer

And then do:

RELOAD;

Then use \q to quit.

Separate Redis for cache

Redis is used widely throughout the application, but some uses are more important than others. Home feeds, list feeds, and Sidekiq queues as well as the streaming API are backed by Redis and that’s important data you wouldn’t want to lose (even though the loss can be survived, unlike the loss of the PostgreSQL database - never lose that!). However, Redis is also used for volatile cache. If you are at a stage of scaling up where you are worried if your Redis can handle everything, you can use a different Redis database for the cache. In the environment, you can specify CACHE_REDIS_URL or individual parts like CACHE_REDIS_HOST, CACHE_REDIS_PORT etc. Unspecified parts fallback to the same values as without the cache prefix.

As far as configuring the Redis database goes, basically you can get rid of background saving to disk, since it doesn’t matter if the data gets lost on restart and you can save some disk I/O on that. You can also add a maximum memory limit and a key eviction policy, for that, see this guide: Using Redis as an LRU cache

Read-replicas

To reduce the load on your Postgresql server, you may wish to setup hot streaming replication (read replica). See this guide for an example. You can make use of the replica in Mastodon in these ways:

• The streaming API server does not issue writes at all, so you can connect it straight to the replica. But it’s not querying the database very often anyway so the impact of this is little.
• Use the Makara driver in the web and sidekiq processes, so that writes go to the primary database, while reads go to the replica. Let’s talk about that.

You will have to edit the config/database.yml file and replace the production section as follows:

production:
  <<: *default
  adapter: postgresql_makara
  prepared_statements: false
  makara:
    id: postgres
    sticky: true
    connections:
      - role: master
        blacklist_duration: 0
        url: postgresql://db_user:db_password@db_host:db_port/db_name
      - role: slave
        url: postgresql://db_user:db_password@db_host:db_port/db_name

Make sure the URLs point to wherever your PostgreSQL servers are. You can add multiple replicas. You could have a locally installed pgBouncer with configuration to connect to two different servers based on database name, e.g. “mastodon” going to the primary, “mastodon_replica” going to the replica, so in the file above both URLs would point to the local pgBouncer with the same user, password, host and port, but different database name. There are many possibilities how this could be setup! For more information on Makara, see their documentation.