Restoring an OpenProject backup

This document describes how to restore a complete backup of OpenProject.

Please look here if you want to restore a deleted project from a backup.

Package-based installation (DEB/RPM)

Assuming you have a backup of all the OpenProject files at hand (see the Backing up guide), here is how you would restore your OpenProject installation from that backup.

As a reference, we will assume you have the following dumps on your server, located in /var/db/openproject/backup:

ubuntu@ip-10-0-0-228:/home/ubuntu# sudo ls -al /var/db/openproject/backup/
total 1680
drwxr-xr-x 2 openproject openproject    4096 Nov 19 21:00 .
drwxr-xr-x 6 openproject openproject    4096 Nov 19 21:00 ..
-rw-r----- 1 openproject openproject 1361994 Nov 19 21:00 attachments-20191119210038.tar.gz
-rw-r----- 1 openproject openproject    1060 Nov 19 21:00 conf-20191119210038.tar.gz
-rw-r----- 1 openproject openproject     126 Nov 19 21:00 git-repositories-20191119210038.tar.gz
-rw-r----- 1 openproject openproject  332170 Nov 19 21:00 postgresql-dump-20191119210038.pgdump
-rw-r----- 1 openproject openproject     112 Nov 19 21:00 svn-repositories-20191119210038.tar.gz

Stop the processes

First, it is a good idea to stop the OpenProject instance:

sudo service openproject stop

Restoring assets

Untar the attachments to their destination:

sudo tar xzf /var/db/openproject/backup/attachments-20191119210038.tar.gz -C /var/db/openproject/files

Untar the configuration files to their destination:

sudo tar xzf /var/db/openproject/backup/conf-20191119210038.tar.gz -C /etc/openproject

If you want to change anything in the configuration, you can also inspect the /etc/openproject folder afterwards and change them accordingly. To go through all configured wizards steps, use the openproject reconfigure option. See the configuration guide for more information.

Untar the repositories to their destination:

sudo tar xzf /var/db/openproject/backup/git-repositories-20191119210038.tar.gz -C /var/db/openproject/git
sudo tar xzf /var/db/openproject/backup/svn-repositories-20191119210038.tar.gz -C /var/db/openproject/svn

Restoring the database

Note: in this section, the <dbusername>, <dbhost> and <dbname> variables that appear below have to be replaced with the values that are contained in the DATABASE_URL setting of your installation.

If you are moving OpenProject to a new server and want to import the backup there, this will be the DATABASE_URL of your new installation on that server.

First, ensure the connection details about your database is the one you want to restore

sudo openproject config:get DATABASE_URL
#=> e.g.: postgres://<dbusername>:<dbpassword>@<dbhost>:<dbport>/<dbname>

Then, to restore the PostgreSQL dump please use the pg_restore command utility. WARNING: The command --clean --if-exists is used and it will drop objects in the database and you will lose all changes in this database! Double-check that the database URL above is the database you want to restore to.

This is necessary since the backups of OpenProject does not clean statements to remove existing options and will lead to duplicate index errors when trying to restore to an existing database. The alternative is to drop/recreate the database manually (see below), if you have the permissions to do so.

sudo pg_restore --clean --if-exists --dbname $(sudo openproject config:get DATABASE_URL) postgresql-dump-20200804094017.pgdump

As the pg_restore tries to apply the username from the dumped database as the owner, you might see errors if you restoring to a database with a different username. In this case, please add --no-owner as a command line argument.

NOTE: If the backup was made in the OpenProject Enterprise-Cloud, please navigate to Changing the database schema from cloud to on-premises

Troubleshooting

Restore fails with something like ‘Error while PROCESSING TOC […] cannot drop constraint’

In this case you will have to drop and re-create the database, and then import it again. If you have access to the postgres user, it’s simply a matter of starting the psql console like this:

sudo su - postgres -c psql

And once in there drop and re-create the database. Ensure that the new database has the correct name and owner. You can get these values from the DATABASE_URL as shown above.

DROP DATABASE openproject; CREATE DATABASE openproject OWNER openproject;

Once done you can exit the psql console by entering \q. Now you can restore the database as seen above.

Restart the OpenProject processes

Finally, restart all your processes as follows:

sudo service openproject restart

Docker-based installation

For Docker-based installations, assuming you have a backup as per the procedure described in the Backing up guide, you simply need to restore files into the correct folders (when using the all-in-one container), or restore the docker volumes (when using the Compose file), then start OpenProject using the normal docker or docker-compose command.

Using docker-compose

Let’s assume you want to restore a database dump given in a file, say openproject.sql.

This assumes that the database container is called compose_db_1. Find out the actual name on your host using docker ps | grep postgres.

If you are using docker-compose this is what you do after you started everything for the first time using docker-compose up -d:

Stop the OpenProject container using docker-compose stop web worker.
Drop the existing, seeded database using docker exec -it compose_db_1 psql -U postgres -c 'drop database openproject;'
If your database doesn’t have an openproject user yet, create it with this command: docker exec -it compose_db_1 psql -U postgres -c 'create user openproject;'
Recreate the database using docker exec -it compose_db_1 psql -U postgres -c 'create database openproject owner openproject;'
Copy the dump onto the container: docker cp openproject.sql compose_db_1:/
Source the dump with psql on the container: docker exec -it compose_db_1 psql -U postgres followed first by \c openproject and then by \i openproject.sql. You can leave this console by entering \q once it’s done.
Delete the dump on the container: docker exec -it compose_db_1 rm openproject.sql
Run the seeder once to perform any migrations docker-compose start seeder
Restart the web and worker processes: docker-compose start web worker
Confirm with docker-compose logs -f that the processes are starting up correctly.

NOTE: If the backup was made in the OpenProject Enterprise-Cloud, please navigate to Changing the database schema from cloud to on-premises

Using the all-in-one container

Given a SQL dump openproject.sql (or a .pgdump file) we can create a new OpenProject container using it with the following steps.

Create the pgdata folder to be mounted in the OpenProject container.
Initialize the database.
Restore the dump.
Start the OpenProject container mounting the pgdata folder.

1) Create the folders to be mounted

First we create the folder to be mounted by our OpenProject container. While we’re at we also create the assets folder which should be mounted too.

mkdir -p /var/lib/openproject/{pgdata,assets}

2) Initialize the database

Next we need to initialize the database.

docker run --rm -v /var/lib/openproject/pgdata:/var/openproject/pgdata -it openproject/openproject:15

As soon as you see Database setup finished. in the container’s output you can kill it by pressing Ctrl + C. It may take a moment to shut down. This then has initialized the database under /var/lib/openproject/pgdata on your docker host.

3) Restore the dump

Now we can restore the database. For this we mount the initialized pgdata folder using the postgres docker container.

docker run --rm -d --name postgres -v /var/lib/openproject/pgdata:/var/lib/postgresql/data postgres:13

Once the container is ready you can copy your SQL dump onto it and start psql.

docker cp openproject.sql postgres:/
docker exec -it postgres psql -U postgres

In psql you then restore dump like this:

DROP DATABASE openproject;
CREATE DATABASE openproject OWNER openproject;

\c openproject
\i openproject.sql

Once this has finished you can quit psql (using \q) and the container (exit).

Importing backups from a package-based installation

If you have a .pgdump file instead, for instance from a backup of a package-based OpenProject installation, the process works almost the same. You still just copy the file into the container as shown above, but then you use pg_restore instead to restore it.

# 1. copy .pgdump file into container
docker cp postgresql-dump-20211119210038.pgdump postgres:/

# 2. delete existing database created in step 2) above
docker exec -it postgres dropdb -U postgres openproject

# 3. import the dump
docker exec -it postgres pg_restore -U postgres postgresql-dump-20211119210038.pgdump

Dump restored

Once the dump is restored you can stop the postgres container using docker stop postgres. Now you have to fix the permissions that were changed by the postgres container so OpenProject can use the files again.

chown -R 102:102 /var/lib/openproject/pgdata

Your pgdata directory is now ready to be mounted by your final OpenProject container.

Restoring attachments

If you also have file attachments to restore you can simply copy them into the attachments folder on the docker host which is mounted into the OpenProject container. For instance:

# 1. extract files
tar -C /var/lib/openproject/assets -xf attachments-20210211090802.tar.gz

# 2. give right permission so `app` user in container can read them
chown -R 1000:1000 /var/lib/openproject/assets

You may need to create the files directory if it doesn’t exist yet.

4) Start OpenProject

Start the container as described in the installation section mounting /var/lib/openproject/pgdata (and /var/lib/openproject/assets/ for attachments).

Changing the database schema from cloud to on-premises

If you want to restore a dump from the OpenProject cloud to your on-premises installation, you will have to change the schema name in the database.

Cloud schemas have a long alpha-numeric name, for instance 123456789_1234567_1234567a_123b_12c3_1234_c2a1a123c123. If you want to use this on-premises you will have to rename that to the default which is public.

Stop OpenProject (service openproject stop) but keep the database up and running.
Connect to your PSQL database (psql $(openproject config:get DATABASE_URL)).
Double check the existing schemas using \dn on the psql console.
If there are indeed 2 schemas, that is an extra schema on top of public, drop the public one and rename the other accordingly.

openproject=# DROP SCHEMA public CASCADE;
DROP SCHEMA
openproject=# ALTER SCHEMA "123456789_1234567_1234567a_123b_12c3_1234_c2a1a123c123" RENAME TO public;
ALTER SCHEMA