A pgbarman docker image with Prometheus metrics

After properly configured this image runs a cron with the cron command every minute. That "executes WAL archiving operations concurrently on a server basis, and this also enforces retention policies on those servers that have:

• retention_policy not empty and valid;
• retention_policy_mode set to auto.

The cron command ensures that WAL streaming is started for those servers that have requested it, by transparently executing the receive-wal command."

The image currently is focused on streaming replication so no rsync streaming or now.

$ docker run -v /path/to/db/configs:/etc/barman.d/ -v /path/to/backupdata/:/var/lib/barman/ decsis/pg-barman

For an full overview of using this image with a server check the example configuration for in the docker-compose.yml. If you run this compose file you will obtain two PostgresSQL servers and a Barman scheduled to make full backups every hour for server1 and every 6 hours to server2. Also a Prometheus server will be available to gather metrics from the exporter included in the image.

docker-compose up

After that Barman will complain in the log:

barman_1        | 2017-03-17 11:54:02,239 [192] barman.server ERROR: replication slot 'barman' doesn't exist. Please execute 'barman receive-wal --create-slot postgreslab1'
barman_1        | 2017-03-17 11:54:02,256 [193] barman.wal_archiver INFO: Synchronous WAL streaming for barman_receive_wal: False
barman_1        | 2017-03-17 11:54:02,257 [193] barman.server ERROR: replication slot 'barman' doesn't exist. Please execute 'barman receive-wal --create-slot postgreslab2'

Just run:

docker exec -u barman -it dockerbarman_barman_1 barman receive-wal --create-slot postgreslab1
docker exec -u barman -it dockerbarman_barman_1 barman receive-wal --create-slot postgreslab2

Give it a minute or so to settle up and then execute the following to make PostgreSQL servers switch to another transaction log file and allows barman to be ready to do backups:

docker exec -u barman -it dockerbarman_barman_1 barman switch-xlog --force all

Prepare the recovery ssh connexion from barman to rec_pg:

docker exec -u barman -it dockerbarman_barman_1 ssh-keygen -t rsa && cat ~/.ssh/id_rsa.pub

Allow the barman's public key just displayed on rec_pg:

docker exec -u root -it dockerbarman_rec_pg_1 sh 
# echo the key to ~/.ssh/authorized_key 
# then chhmod 0600 ~/.ssh/authorized_key

Finally execute the backups:

docker exec -u barman -it dockerbarman_barman_1 barman backup all

At any point you can view the status at http://localhost:8000

IMPORTANT: Take this just as a quick overview of the image funcionality. Please do read the barman manual and the PostgreSQL documentation to fully understand what is going on.

All ENV variables that starts with BARMAN_ will be converted read and converted to the correct format needed for barman.conf file. Example BARMAN_MINIMUM_REDUNDANCY=1 will be inserted as minimum_redundancy=1. Full list of options at pgbarman's manual.

Alternatively it is possible to mount a volume file in $BARMAN_BARMAN_HOME/.barman.conf that will override the base config file.

For server configurations provide a file per server in /etc/barman.d/ as stated in the barman manual.

Additionaly you can pass a extra configuration in these files for scheduling backups:

#:backup_cron = 0 10 * * *
#:backup_recovery_ssh_cmd = "ssh root@rec_pg"

backup_cron

The image runs a every minute script that maintains (insert, update and delete) the crontab gathering this property from the available *.conf files. The resulting crontab for the previous example:

MAILTO=""
BARMAN_LOG_FILE=/var/log/barman.log
BARMAN_PRE_BACKUP_SCRIPT=/opt/barman/scripts/pre_backup.sh
BARMAN_POST_BACKUP_SCRIPT=/opt/barman/scripts/post_backup.sh
BARMAN_BARMAN_HOME=/var/lib/barman
BARMAN_CONFIGURATION_FILES_DIRECTORY=/etc/barman.d
* * * * * barman cron
* * * * * /opt/barman/scripts/backup_scheduler.sh
0 10 * * * barman backup postgreslab3

backup_recovery_ssh_cmd

If this property is present after each backup of the corresponding database a recovery attempt will be performed and its duration registered as a Prometheus metric.

Currently this only works with the companion image https://hub.docker.com/r/decsis/postgres-ssh/ .

Important: Please notice that this additional properties are prefixed by #: to avoid barman complaining about an unknown property. The : is there to make clear this is not just a comment.

The image ships with a Prometheus exporter in the form of a Python script. After the container starts the metrics should be available at http://localhost:8000.

It will generate the following metrics:

• barman_status Status as given by barman check command (0 is OK everything else is bad)
• barman_last_backup_size_bytes Last backup size in bytes
• barman_expected_redundancy Number of expected backups as defined in config minimum_redundancy
• barman_database_size_bytes Database size in bytes
• barman_backup_duration_seconds Backup duration in seconds
• barman_recovery_duration_seconds Recovery duration in seconds
• barman_recovery_status Recovery status per database
• barman_backup_window_seconds Backup window covered by all existing backups
• barman_last_backup_age_seconds Last backup age
• barman_current_redundancy Number of existing backups

A sample scrape config is available in example/prometheus_config.yml