MatrixSynapse/docs/development/database_schema.md

Synapse database schema files

Synapse's database schema is stored in the synapse.storage.schema module.

Logical databases

Synapse supports splitting its datastore across multiple physical databases (which can be useful for large installations), and the schema files are therefore split according to the logical database they apply to.

At the time of writing, the following "logical" databases are supported:

• state - used to store Matrix room state (more specifically, state_groups, their relationships and contents).
• main - stores everything else.

Additionally, the common directory contains schema files for tables which must be present on all physical databases.

Synapse schema versions

Synapse manages its database schema via "schema versions". These are mainly used to help avoid confusion if the Synapse codebase is rolled back after the database is updated. They work as follows:

• The Synapse codebase defines a constant synapse.storage.schema.SCHEMA_VERSION which represents the expectations made about the database by that version. For example, as of Synapse v1.36, this is 59.

• The database stores a "compatibility version" in schema_compat_version.compat_version which defines the SCHEMA_VERSION of the oldest version of Synapse which will work with the database. On startup, if compat_version is found to be newer than SCHEMA_VERSION, Synapse will refuse to start.

  Synapse automatically updates this field from synapse.storage.schema.SCHEMA_COMPAT_VERSION.

• Whenever a backwards-incompatible change is made to the database format (normally via a delta file), synapse.storage.schema.SCHEMA_COMPAT_VERSION is also updated so that administrators can not accidentally roll back to a too-old version of Synapse.

Generally, the goal is to maintain compatibility with at least one or two previous releases of Synapse, so any substantial change tends to require multiple releases and a bit of forward-planning to get right.

As a worked example: we want to remove the room_stats_historical table. Here is how it might pan out.

1. Replace any code that reads from room_stats_historical with alternative implementations, but keep writing to it in case of rollback to an earlier version. Also, increase synapse.storage.schema.SCHEMA_VERSION. In this instance, there is no existing code which reads from room_stats_historical, so our starting point is:

   v1.36.0: SCHEMA_VERSION=59, SCHEMA_COMPAT_VERSION=59

2. Next (say in Synapse v1.37.0): remove the code that writes to room_stats_historical, but don’t yet remove the table in case of rollback to v1.36.0. Again, we increase synapse.storage.schema.SCHEMA_VERSION, but because we have not broken compatibility with v1.36, we do not yet update SCHEMA_COMPAT_VERSION. We now have:

   v1.37.0: SCHEMA_VERSION=60, SCHEMA_COMPAT_VERSION=59.

3. Later (say in Synapse v1.38.0): we can remove the table altogether. This will break compatibility with v1.36.0, so we must update SCHEMA_COMPAT_VERSION accordingly. There is no need to update synapse.storage.schema.SCHEMA_VERSION, since there is no change to the Synapse codebase here. So we end up with:

   v1.38.0: SCHEMA_VERSION=60, SCHEMA_COMPAT_VERSION=60.

If in doubt about whether to update SCHEMA_VERSION or not, it is generally best to lean towards doing so.

Full schema dumps

In the full_schemas directories, only the most recently-numbered snapshot is used (54 at the time of writing). Older snapshots (eg, 16) are present for historical reference only.

Building full schema dumps

If you want to recreate these schemas, they need to be made from a database that has had all background updates run.

To do so, use scripts-dev/make_full_schema.sh. This will produce new full.sql.postgres and full.sql.sqlite files.

Ensure postgres is installed, then run:

./scripts-dev/make_full_schema.sh -p postgres_username -o output_dir/

NB at the time of writing, this script predates the split into separate state/main databases so will require updates to handle that correctly.

Delta files

Delta files define the steps required to upgrade the database from an earlier version. They can be written as either a file containing a series of SQL statements, or a Python module.

Synapse remembers which delta files it has applied to a database (they are stored in the applied_schema_deltas table) and will not re-apply them (even if a given file is subsequently updated).

Delta files should be placed in a directory named synapse/storage/schema/<database>/delta/<version>/. They are applied in alphanumeric order, so by convention the first two characters of the filename should be an integer such as 01, to put the file in the right order.

SQL delta files

These should be named *.sql, or — for changes which should only be applied for a given database engine — *.sql.posgres or *.sql.sqlite. For example, a delta which adds a new column to the foo table might be called 01add_bar_to_foo.sql.

Note that our SQL parser is a bit simple - it understands comments (-- and /*...*/), but complex statements which require a ; in the middle of them (such as CREATE TRIGGER) are beyond it and you'll have to use a Python delta file.

Python delta files

For more flexibility, a delta file can take the form of a python module. These should be named *.py. Note that database-engine-specific modules are not supported here – instead you can write if isinstance(database_engine, PostgresEngine) or similar.

A Python delta module should define either or both of the following functions:

import synapse.config.homeserver
import synapse.storage.engines
import synapse.storage.types


def run_create(
    cur: synapse.storage.types.Cursor,
    database_engine: synapse.storage.engines.BaseDatabaseEngine,
) -> None:
    """Called whenever an existing or new database is to be upgraded"""
    ...

def run_upgrade(
    cur: synapse.storage.types.Cursor,
    database_engine: synapse.storage.engines.BaseDatabaseEngine,
    config: synapse.config.homeserver.HomeServerConfig,
) -> None:
    """Called whenever an existing database is to be upgraded."""
    ...

Boolean columns

Boolean columns require special treatment, since SQLite treats booleans the same as integers.

Any new boolean column must be added to the BOOLEAN_COLUMNS list in synapse/_scripts/synapse_port_db.py. This tells the port script to cast the integer value from SQLite to a boolean before writing the value to the postgres database.

event_id global uniqueness

event_id's can be considered globally unique although there has been a lot of debate on this topic in places like MSC2779 and MSC2848 which has no resolution yet (as of 2022-09-01). There are several places in Synapse and even in the Matrix APIs like GET /_matrix/federation/v1/event/{eventId} where we assume that event IDs are globally unique.

When scoping event_id in a database schema, it is often nice to accompany it with room_id (PRIMARY KEY (room_id, event_id) and a FOREIGN KEY(room_id) REFERENCES rooms(room_id)) which makes flexible lookups easy. For example it makes it very easy to find and clean up everything in a room when it needs to be purged (no need to use sub-select query or join from the events table).

A note on collisions: In room versions 1 and 2 it's possible to end up with two events with the same event_id (in the same or different rooms). After room version 3, that can only happen with a hash collision, which we basically hope will never happen (SHA256 has a massive big key space).

Worked examples of gradual migrations

Some migrations need to be performed gradually. A prime example of this is anything which would need to do a large table scan — including adding columns, indices or NOT NULL constraints to non-empty tables — such a migration should be done as a background update where possible, at least on Postgres. We can afford to be more relaxed about SQLite databases since they are usually used on smaller deployments and SQLite does not support the same concurrent DDL operations as Postgres.

We also typically insist on having at least one Synapse version's worth of backwards compatibility, so that administrators can roll back Synapse if an upgrade did not go smoothly.

This sometimes results in having to plan a migration across multiple versions of Synapse.

This section includes an example and may include more in the future.

Transforming a column into another one, with NOT NULL constraints

This example illustrates how you would introduce a new column, write data into it based on data from an old column and then drop the old column.

We are aiming for semantic equivalence to:

ALTER TABLE mytable ADD COLUMN new_column INTEGER;
UPDATE mytable SET new_column = old_column * 100;
ALTER TABLE mytable ALTER COLUMN new_column ADD CONSTRAINT NOT NULL;
ALTER TABLE mytable DROP COLUMN old_column;

Synapse version N

SCHEMA_VERSION = S
SCHEMA_COMPAT_VERSION = ... # unimportant at this stage

Invariants:

1. old_column is read by Synapse and written to by Synapse.

Synapse version N + 1

SCHEMA_VERSION = S + 1
SCHEMA_COMPAT_VERSION = ... # unimportant at this stage

Changes: 1.

ALTER TABLE mytable ADD COLUMN new_column INTEGER;

Invariants:

1. old_column is read by Synapse and written to by Synapse.
2. new_column is written to by Synapse.

Notes:

1. new_column can't have a NOT NULL NOT VALID constraint yet, because the previous Synapse version did not write to the new column (since we haven't bumped the SCHEMA_COMPAT_VERSION yet, we still need to be compatible with the previous version).

Synapse version N + 2

SCHEMA_VERSION = S + 2
SCHEMA_COMPAT_VERSION = S + 1 # this signals that we can't roll back to a time before new_column existed

Changes:

1. On Postgres, add a NOT VALID constraint to ensure new rows are compliant. SQLite does not have such a construct, but it would be unnecessary anyway since there is no way to concurrently perform this migration on SQLite.

   ALTER TABLE mytable ADD CONSTRAINT CHECK new_column_not_null (new_column IS NOT NULL) NOT VALID;
   

2. Start a background update to perform migration: it should gradually run e.g.

   UPDATE mytable SET new_column = old_column * 100 WHERE 0 < mytable_id AND mytable_id <= 5;
   

   This background update is technically pointless on SQLite, but you must schedule it anyway so that the portdb script to migrate to Postgres still works.
3. Upon completion of the background update, you should run VALIDATE CONSTRAINT on Postgres to turn the NOT VALID constraint into a valid one.

   ALTER TABLE mytable VALIDATE CONSTRAINT new_column_not_null;
   

   This will take some time but does NOT hold an exclusive lock over the table.

Invariants:

1. old_column is read by Synapse and written to by Synapse.
2. new_column is written to by Synapse and new rows always have a non-NULL value in this field.

Notes:

1. If you wish, you can convert the CHECK (new_column IS NOT NULL) to a NOT NULL constraint free of charge in Postgres by adding the NOT NULL constraint and then dropping the CHECK constraint, because Postgres can statically verify that the NOT NULL constraint is implied by the CHECK constraint without performing a table scan.
2. It might be tempting to make version N + 2 redundant by moving the background update to N + 1 and delaying adding the NOT NULL constraint to N + 3, but that would mean the constraint would always be validated in the foreground in N + 3. Whereas if the N + 2 step is kept, the migration in N + 3 would be fast in the happy case.

Synapse version N + 3

SCHEMA_VERSION = S + 3
SCHEMA_COMPAT_VERSION = S + 1 # we can't roll back to a time before new_column existed

Changes:

1. (Postgres) Update the table to populate values of new_column in case the background update had not completed. Additionally, VALIDATE CONSTRAINT to make the check fully valid.

   -- you ideally want an index on `new_column` or e.g. `(new_column) WHERE new_column IS NULL` first, or perhaps you can find a way to skip this if the `NOT NULL` constraint has already been validated.
   UPDATE mytable SET new_column = old_column * 100 WHERE new_column IS NULL;
   
   -- this is a no-op if it already ran as part of the background update
   ALTER TABLE mytable VALIDATE CONSTRAINT new_column_not_null;
   

2. (SQLite) Recreate the table by precisely following the 12-step procedure for SQLite table schema changes. During this table rewrite, you should recreate new_column as NOT NULL and populate any outstanding NULL values at the same time. Unfortunately, you can't drop old_column yet because it must be present for compatibility with the Postgres schema, as needed by portdb. (Otherwise you could do this all in one go with SQLite!)

Invariants:

1. old_column is written to by Synapse (but no longer read by Synapse!).
2. new_column is read by Synapse and written to by Synapse. Moreover, all rows have a non-NULL value in this field, as guaranteed by a schema constraint.

Notes:

1. We can't drop old_column yet, or even stop writing to it, because that would break a rollback to the previous version of Synapse.
2. Application code can now rely on new_column being populated. The remaining steps are only motivated by the wish to clean-up old columns.

Synapse version N + 4

SCHEMA_VERSION = S + 4
SCHEMA_COMPAT_VERSION = S + 3 # we can't roll back to a time before new_column was entirely non-NULL

Invariants:

1. old_column exists but is not written to or read from by Synapse.
2. new_column is read by Synapse and written to by Synapse. Moreover, all rows have a non-NULL value in this field, as guaranteed by a schema constraint.

Notes:

1. We can't drop old_column yet because that would break a rollback to the previous version of Synapse.
   TODO: It may be possible to relax this and drop the column straight away as long as the previous version of Synapse detected a rollback occurred and stopped attempting to write to the column. This could possibly be done by checking whether the database's schema compatibility version was S + 3.

Synapse version N + 5

SCHEMA_VERSION = S + 5
SCHEMA_COMPAT_VERSION = S + 4 # we can't roll back to a time before old_column was no longer being touched

Changes: 1.

ALTER TABLE mytable DROP COLUMN old_column;