Last updated 12 December 2016
Table of Contents
Backups of your data are crucial to any application, not just for recovering in case of a catastrophe, but also for testing, setting up staging environments, and migrating your data. Every Heroku Postgres database comes with a behind-the-scenes Continuous Protection mechanism for disaster recovery, as well as optional “logical” backups under your direct control.
This article explains how to take manual and scheduled backups, show your existing backups, restore backups, and transfer data directly between two databases.
For all production Heroku Postgres plans, using continuous protection is the recommended path for data backup and recovery.
Creating a backup
pg:backups operates against your primary database,
identified by the
DATABASE_URL config var:
$ heroku pg:backups:capture --app sushi Hit Ctrl-C at any time to stop watching progress; the backup will continue running. Stop a running backup with heroku pg:backups:cancel. HEROKU_POSTGRESQL_BLACK (DATABASE_URL) ----backup---> b251 Running... done
If you have multiple databases on your application, you can choose which one to backup by specifying the database name:
$ heroku pg:backups:capture HEROKU_POSTGRESQL_PINK Hit Ctrl-C at any time to stop watching progress; the backup will continue running. Stop a running backup with heroku pg:backups:cancel. HEROKU_POSTGRESQL_PINK ----backup---> b252 Running... done
You can use the flag
--verbose to see logs as your backup progresses.
If you need to stop a backup for any reason, use the
$ heroku pg:backups:cancel Canceled backup b252
Manual backup retention limits
There is a limit to the number of manual backups that you can retain. That number is based on your database plan.
If you’ve reached this limit and need to take additional backups, the capture command will automatically expire the oldest backup before capturing a new one.
Capturing a backup adds some load on your database for the duration of the backup. How this impacts your application will vary with the size of your database and the nature of the app. Consider taking backups on a follower if there is a significant impact from running them on the master.
In addition to manually triggered backups, you can schedule regular automatic backups. These will run daily against the specified database.
$ heroku pg:backups:schedule DATABASE_URL --at '02:00 America/Los_Angeles' --app sushi
--at option uses a 24 hour clock to indicate the hour of the day that you want the backup to be taken. The
--at option also accepts a timezone in either the full TZ format (America/Los_Angeles) or the abbreviation (PST) but we recommend you use the full TZ format.
--at option is required when scheduling a backup.
If you upgrade the database from hobby tier to production tier, the schedule will be lost.
When a backup is restored to a new database, the schedule from the original database does not get restored, a new schedule needs to be created.
To stop regular backups, use
$ heroku pg:backups:unschedule DATABASE_URL --app sushi
To view current schedules for your app, use:
$ heroku pg:backups:schedules --app sushi Current backup schedules: RED: daily at 2:00 (America/Los_Angeles)
Scheduled backups retention limits
Scheduled backups have a different retention policy as compared to the manual backups. This policy is also based on the database plan. For all plans, a daily backup is retained for the last 7 days. That means that 7 backups will exist, one for each of the last 7 days. A weekly backup means that only one backup is saved over a 7 day period. A monthly backup means that only 1 backup is saved over the course of a month. Based on current limits, for example, a Premium-0 would have 12 monthly backups, one for each of the last 12 months.
|Plan||Weekly Backups Retained||Monthly Backup Retained|
|Hobby-Dev||1 week||0 months|
|Hobby-Basic||1 week||0 months|
|Standard-*||4 weeks||0 months|
|Premium-*||8 weeks||12 months|
|Enterprise||8 weeks||12 months|
Downloading your backups
You can create a publicly accessible backup URL with the
command. When the command is specified without a backup id, the latest available backup URL will be returned. This is useful for exporting your data outside of Heroku Postgres:
$ heroku pg:backups:url b001 --app sushi The following URL will expire at 2015-04-07 18:35:50 +0000: "http://firstname.lastname@example.org/b004.dump?AWSAccessKeyId=ABCD1234&Expires=1289261668&Signature=3mMBeKISewgEUDT%2FL5mRz4EYS4M%3D"
The URL is public but is not easily guessable. It expires after 60 minutes.
pg:backups:url command will output when the time the URL will expire along with the URL itself. If you plan on piping the URL to other commands,
pg:backups:url will refrain from adding the expiration information to the output so that all subsequent commands will just have the URL. For example, if I just wanted to list the URL in the terminal without the extra information:
$ heroku pg:backups:url --app sushi | cat http://email@example.com/b004.dump?AWSAccessKeyId=ABCD1234&Expires=1289261668&Signature=3mMBeKISewgEUDT%2FL5mRz4EYS4M%3D
To download your backup via the command line, you can use the
$ heroku pg:backups:download
Please see our documentation on importing and exporting Heroku Postgres databases with PGBackups for more information.
Checking backup status
To list your backups and your continuous protection status, you can run the following:
$ heroku pg:backups --app sushi === Backups ID Backup Time Status Size Database ---- ------------------------- ---------------------------------- ------ -------- b013 2015-03-18 19:03:16 +0000 Running IVORY b011 2015-02-18 17:55:38 +0000 Finished 2015-02-18 17:55:39 +0000 1.9GB IVORY b010 2015-02-17 19:14:43 +0000 Finished 2015-02-17 19:14:48 +0000 1.9GB IVORY b004 2015-02-11 19:00:55 +0000 Finished 2015-02-17 19:14:48 +0000 1.9GB IVORY ==== Restores ID Restore Time Status Size Database ---- ------------------------- ---------------------------------- ------ -------- r002 2015-03-16 17:33:19 +0000 Finished 2015-03-16 17:33:19 +0000 1.9GB IVORY r001 2015-03-15 12:13:44 +0000 Failed 2015-03-15 12:13:47 +0000 1.7GB IVORY
The continuous protection status shows how your database is covered in case disaster recovery is necessary. The last base backup will be the starting point. To restore your database, all WAL since that point will be replayed until the last WAL segment archived. Any data up to the “Last WAL archived” time is safely backed up.
To get more detailed information about a given backup, use the info command:
$ heroku pg:backups:info b017 --app sushi === Backup info: b017 Database: HEROKU_POSTGRESQL_IVORY Started: 2013-06-24 17:11.28 UTC Status: Running Type: Manual Size: 1.2GB Schema: 0bff3ac
With the optional
--verbose flag, you can also see the logs of the
backup in question. If the backup is still running, the command will
print the ongoing logs until the backup finishes or you cancel the command
PG Backups stores backups in a compressed binary format, and the backups include commands to recreate indexes instead of storing the indexes themselves. As a result, backups will be much smaller in size than your database’s current size on disk from
You can manually delete a backup to remove obsolete backups or make room for new captures. Use the backup ID to specify which backup to remove.
$ heroku pg:backups:delete b101 --app foo Deleting b101... done
To restore a backup, use the restore command:
$ heroku pg:backups:restore b101 DATABASE_URL --app sushi
This will restore backup id
b101 to the specified database URL in the app
sushi. Note you can omit the backup id to simply restore the latest backup.
You can also restore from a backup on another app (from sushi app to sushi-staging app):
$ heroku pg:backups:restore sushi::b101 DATABASE_URL --app sushi-staging
Or from a publicly accessible URL:
$ heroku pg:backups:restore 'https://s3.amazonaws.com/me/items/mydb.dump' DB_URL -a sushi
Over time, a database will grow on disk due to table bloat and unused index data. As a result, restoring a backup into a new database will almost always result in a reduction in overall disk size as reported by
Unlike with the previous
pgbackups commands, you cannot restore a partial backup into an existing database. When you run
pg:backups:restore, all data is deleted from the target database before restoring the backup.
Direct database-to-database copies
In addition to backups and restores, pg:backups also provides the ability to transfer data directly between databases.
To perform such a transfer, run the following:
$ heroku pg:copy COBALT GREEN --app sushi
This would copy all data from the
COBALT database to the
GREEN database in the
You can also transfer directly from a database on another app:
$ heroku pg:copy sushi::ORANGE GREEN --app sushi-staging
This would copy data from the
ORANGE database of the
sushi app to the
GREEN database in
sushi-staging. This could be used to copy production data into a staging app for testing purposes.
Heroku Postgres needs to connect to your database using your credentials in order to capture a backup. This connection counts against your plan connection limit. You can identify connections made by Heroku Postgres by running the following command from your terminal:
$ heroku pg:psql -c "select * from pg_stat_activity where application_name = 'heroku-postgres-backups'"