opper/backopper

Backup solution for apps on Serverpilot

Backopper

Backopper is a small tool created in go to assist with the creation of backups for all the projects we got going on. Basically, it runs every x amount of time for each app and it creates the backups somewhere on disk (normally in /opt/backups/$APP). It also syncs the backups to a bucket in S3 and posts an archive of media files to a backup server.

In-depth explanation

There's 2 main functionalities to this tool: fetching a list of apps and creating the actual backups.

Fetching the list of apps

This is done via an API request to the a certain application. A response for this request would look like this:

[
    {
        "name": "project_1",
        "frequency": "daily",
        "db_engine": "mysql",
        "project_type": "wordpress"
    },
    {
        "name": "project_2",
        "frequency": "weekly",
        "db_engine": "postgres",
        "project_type": "laravel"
    }
]

There's 2 main things for each application that come in this response:

• The name of the application (which should match the name of the folder (if it's a serverpilot app) on disk (like /srv/users/serverpilot/apps/project_1))
• The frequency of the backups

It is based on jasonlvhit/gocron meaning that it will not create actual entries in the crontab but it will schedule the jobs necessary itself. Makes things a bit cleaner (though a bit harder to debug)

Creating the actual backup

When the actual backup handler is run, several things happen:

• First, it looks for the .env file inside that app's root folder (hence the importance of the name of the app matching its location on disk).
• Once it found it, it evaluates that .env file so that the script has access to all the database credentials.
• After that, it composes a mysqldump or pg_dump command based on those parameters. An example of that would look like so: mysqldump --user="$USER" --password="$PASSWORD" $DB_NAME | gzip > $BACKUPS_LOCATION/$UNIX_TIMESTAMP.sql.gz. So, in the end, a new file that will have the name of the unix timestamp at the time of creating the backup will be created under /opt/backups/$APP. I know that having the password in 'plain text' while executing the backup is not secure but this is the best solution without overengineering everything.
• If the database dump is successful, it syncs a copy to an S3 bucket with a key like so: $PROJECT_NAME/$ENVIRONMENT/backup_$TIMESTAMP.sql.gz if it's a MySQL dump or $PROJECT_NAME/$ENVIRONMENT/backup_$TIMESTAMP.tar if it's a PostgreSQL dump.
• Regardless if the database dump is successful or not, it will attempt to create a media backup (generally things like pictures that users upload, documents, etc). It is quite straightforward in the sense that it will simply run tar in the following format: tar -czf $TEMPORARY_MEDIA_LOCATION -C $MEDIA_FOLDER $FOLDER_TO_BACK_UP. A temporary location is used because the media backup can be quite heavy and it is not stored on the server itself (more on that later). The -C flag is used to change the directory location to wherever the media folder is located, but not the actual folder to back up itself, that is $FOLDER_TO_BACK_UP. It is a bit complex in the sense that it changes dirs to the parent folder to back up and then it performs the tar command but this is the most reliable way I could find of making it work.
• If the media dump is successful, it is then sent via scp to the media server. A command like the following will be executed: scp -P$PORT $TEMPORARY_MEDIA_LOCATION $USER@$HOST:$MEDIA_BACKUPS_PATH.