Add support for backing up / importing all Postgres databases/users

CHANGELOG.md

@@ -1,3 +1,17 @@
+# 2019-10-05
+
+## Improved Postgres upgrading/importing
+
+Postgres [upgrading](docs/maintenance-postgres.md#upgrading-postgresql) and [importing](docs/importing-postgres.md) have been improved to add support for multiple databases and roles.
+
+Previously, the playbook would only take care of the `homeserver` database and `synapse` user.
+We now back up and restore all databases and users on the Postgres server.
+
+For now, the playbook only uses that one database (`homeserver`) and that one single user (`synapse`), so it's all the same.
+However, in the future, additional components besides Synapse may also make use the Postgres database server.
+One such example is the [matrix-appservice-slack](https://github.com/matrix-org/matrix-appservice-slack) bridge, which strongly encourages use of Postgres in its v1.0 release. We are yet to upgrade to it.
+
+
 # 2019-10-04
 
 ## Postgres 12 support

docs/importing-postgres.md

@@ -12,6 +12,8 @@ If your database name differs, be sure to change `matrix_postgres_db_name` to yo
 
 The playbook supports importing Postgres dump files in **text** (e.g. `pg_dump > dump.sql`) or **gzipped** formats (e.g. `pg_dump | gzip -c > dump.sql.gz`).
 
+Importing multiple databases (as dumped by `pg_dumpall`) is also supported.
+
 Before doing the actual import, **you need to upload your Postgres dump file to the server** (any path is okay).
 
 

docs/maintenance-postgres.md

@@ -40,16 +40,18 @@ To make a back up of the current PostgreSQL database, make sure it's running and
 ```bash
 docker run \
 --rm \
---network matrix \
+--network=matrix \
 --env-file=/matrix/postgres/env-postgres-psql \
 postgres:12.0-alpine \
-pg_dump -h matrix-postgres \
+pg_dumpall -h matrix-postgres \
 | gzip -c \
 > /postgres.sql.gz
 ```
 
 If you are using an [external Postgres server](configuring-playbook-external-postgres.md), the above command will not work, because the credentials file (`/matrix/postgres/env-postgres-psql`) is not available.
 
+Restoring a backup made this way can be done by [importing it](importing-postgres.md).
+
 
 ## Upgrading PostgreSQL
 
@@ -64,7 +66,7 @@ This playbook can upgrade your existing Postgres setup with the following comman
 
 	ansible-playbook -i inventory/hosts setup.yml --tags=upgrade-postgres
 
-**The old Postgres data directory is backed up** automatically, by renaming to `/matrix/postgres-auto-upgrade-backup`.
+**The old Postgres data directory is backed up** automatically, by renaming it to `/matrix/postgres-auto-upgrade-backup`.
 To rename to a different path, pass some extra flags to the command above, like this: `--extra-vars="postgres_auto_upgrade_backup_data_path=/another/disk/matrix-postgres-before-upgrade"`
 
 The auto-upgrade-backup directory stays around forever, until you **manually decide to delete it**.
@@ -72,5 +74,4 @@ The auto-upgrade-backup directory stays around forever, until you **manually dec
 As part of the upgrade, the database is dumped to `/tmp`, an upgraded and empty Postgres server is started, and then the dump is restored into the new server.
 To use a different directory for the dump, pass some extra flags to the command above, like this: `--extra-vars="postgres_dump_dir=/directory/to/dump/here"`
 
-**ONLY one database is migrated** (the one specified in `matrix_postgres_db_name`, named `homeserver` by default).
+**All databases, roles, etc. on the Postgres server are migrated**.
-If you've created other databases in that database instance (something this playbook never does and never advises), data will be lost.

roles/matrix-postgres/tasks/import_postgres.yml

@@ -56,6 +56,10 @@
     msg: "Could not find existing Postgres installation"
   when: "not matrix_postgres_detected_existing|bool"
 
+# Starting the database container had automatically created the default
+# role (`matrix_postgres_connection_username`) and database (`matrix_postgres_db_name`).
+# The dump most likely contains those same entries and would try to re-create them, leading to errors.
+# We need to skip over those lines.
 - name: Generate Postgres database import command
   set_fact:
     matrix_postgres_import_command: >-
@@ -67,13 +71,22 @@
       -v {{ server_path_postgres_dump }}:/{{ server_path_postgres_dump|basename }}:ro
       --entrypoint=/bin/sh
       {{ matrix_postgres_docker_image_latest }}
-      -c 'cat /{{ server_path_postgres_dump|basename }} |
+      -c "cat /{{ server_path_postgres_dump|basename }} |
       {{ 'gunzip |' if server_path_postgres_dump.endswith('.gz') else '' }}
-      psql -v ON_ERROR_STOP=1 -h matrix-postgres'
+      grep -vE '^CREATE ROLE {{ matrix_postgres_connection_username }}' |
+      grep -vE '^CREATE DATABASE {{ matrix_postgres_db_name }}' |
+      psql -v ON_ERROR_STOP=1 -h matrix-postgres"
 
+# This is a hack.
+# See: https://ansibledaily.com/print-to-standard-output-without-escaping/
+#
+# We want to run `debug: msg=".."`, but that dumps it as JSON and escapes double quotes within it,
+# which ruins the command (`matrix_postgres_import_command`)
 - name: Note about Postgres importing alternative
-  debug:
+  set_fact:
-    msg: >-
+    dummy: true
+  with_items:
+    - >-
         Importing Postgres database using the following command: `{{ matrix_postgres_import_command }}`.
         If this crashes, you can stop Postgres (`systemctl stop matrix-postgres`),
         delete its existing data (`rm -rf {{ matrix_postgres_data_path }}/*`), start it again (`systemctl start matrix-postgres`)

roles/matrix-postgres/tasks/upgrade_postgres.yml

@@ -67,14 +67,21 @@
   delegate_to: 127.0.0.1
   become: false
 
+# We dump all databases, roles, etc.
+#
+# Because we'll be importing into a new container which initializes the default
+# role (`matrix_postgres_connection_username`) and database (`matrix_postgres_db_name`) by itself on startup,
+# we need to remove these from the dump, or we'll get errors saying these already exist.
 - name: Perform Postgres database dump
-  command: |
+  command: >-
-    /usr/bin/docker run --rm --name matrix-postgres-dump \
+    /usr/bin/docker run --rm --name matrix-postgres-dump
-    --user={{ matrix_user_uid }}:{{ matrix_user_gid }} \
+    --user={{ matrix_user_uid }}:{{ matrix_user_gid }}
-    --network={{ matrix_docker_network }} \
+    --network={{ matrix_docker_network }}
-    --env-file={{ matrix_postgres_base_path }}/env-postgres-psql \
+    --env-file={{ matrix_postgres_base_path }}/env-postgres-psql
-    -v {{ postgres_dump_dir }}:/out \
+    --entrypoint=/bin/sh
-    {{ matrix_postgres_detected_version_corresponding_docker_image }} pg_dump -h matrix-postgres {{ matrix_postgres_db_name }} -f /out/{{ postgres_dump_name }}
+    -v {{ postgres_dump_dir }}:/out
+    {{ matrix_postgres_detected_version_corresponding_docker_image }}
+    -c "pg_dumpall -h matrix-postgres > /out/{{ postgres_dump_name }}"
 
 - name: Ensure matrix-postgres is stopped
   service:
@@ -102,15 +109,43 @@
   delegate_to: 127.0.0.1
   become: false
 
+# Starting the database container had automatically created the default
+# role (`matrix_postgres_connection_username`) and database (`matrix_postgres_db_name`).
+# The dump most likely contains those same entries and would try to re-create them, leading to errors.
+# We need to skip over those lines.
+- name: Generate Postgres database import command
+  set_fact:
+    matrix_postgres_import_command: >-
+      /usr/bin/docker run --rm --name matrix-postgres-import
+      --user={{ matrix_user_uid }}:{{ matrix_user_gid }}
+      --cap-drop=ALL
+      --network={{ matrix_docker_network }}
+      --env-file={{ matrix_postgres_base_path }}/env-postgres-psql
+      --entrypoint=/bin/sh
+      -v {{ postgres_dump_dir }}:/in:ro
+      {{ matrix_postgres_docker_image_latest }}
+      -c "cat /in/{{ postgres_dump_name }} |
+      grep -vE '^CREATE ROLE {{ matrix_postgres_connection_username }}' |
+      grep -vE '^CREATE DATABASE {{ matrix_postgres_db_name }}' |
+      psql -v ON_ERROR_STOP=1 -h matrix-postgres"
+
+# This is a hack.
+# See: https://ansibledaily.com/print-to-standard-output-without-escaping/
+#
+# We want to run `debug: msg=".."`, but that dumps it as JSON and escapes double quotes within it,
+# which ruins the command (`matrix_postgres_import_command`)
+- name: Note about Postgres importing
+  set_fact:
+    dummy: true
+  with_items:
+    - >-
+        Importing Postgres database using the following command: `{{ matrix_postgres_import_command }}`.
+        If this crashes, you can stop Postgres (`systemctl stop matrix-postgres`),
+        delete the new database data (`rm -rf {{ matrix_postgres_data_path }}`)
+        and restore the automatically-made backup (`mv {{ postgres_auto_upgrade_backup_data_path }} {{ matrix_postgres_data_path }}`).
+
 - name: Perform Postgres database import
-  command: |
+  command: "{{ matrix_postgres_import_command }}"
-    /usr/bin/docker run --rm --name matrix-postgres-import \
-    --user={{ matrix_user_uid }}:{{ matrix_user_gid }} \
-    --cap-drop=ALL \
-    --network={{ matrix_docker_network }} \
-    --env-file={{ matrix_postgres_base_path }}/env-postgres-psql \
-    -v {{ postgres_dump_dir }}:/in:ro \
-    {{ matrix_postgres_docker_image_latest }} psql -v ON_ERROR_STOP=1 -h matrix-postgres -f /in/{{ postgres_dump_name }}
 
 - name: Delete Postgres database dump file
   file: