Skip to content

Latest commit

 

History

History
242 lines (186 loc) · 6.56 KB

README.md

File metadata and controls

242 lines (186 loc) · 6.56 KB
Raw

setup_barman

This role is for configuring Barman backups on Postgres nodes.

Requirements

Following are the requirements of this role.

  1. Ansible
  2. edb_devops.edb_postgres -> setup_repo role for setting the repository on the systems.

Role Variables

When executing the role via ansible these are the required variables:

  • pg_version

Postgres Versions supported are: 10, 11, 12, 13 and 14

  • pg_type

Database Engine supported are: PG and EPAS

These and other variables can be assigned in the pre_tasks definition of the section: How to include the setup_barman role in your Playbook

The rest of the variables can be configured and are available in the:

Below is the documentation of the rest of the main variables:

barman_pg_user

Dedicated Postgres user excuting barman queries. Default: barman

Example:

barman_pg_user: 'barman'

Host Variables

Below are the host variables defined in the inventory file, for each Postgres node we want to backup with Barman.

barman

Enable Barman backups for the host. Default: false

Example:

barman: yes

barman_server_private_ip

Barman server private IP address. Default: None

Example:

barman_server_private_ip: 10.0.0.123

barman_backup_method

Backup method. Can be:

  • postgres for backups based on pg_basebackup using Streaming Replication protocol.
  • rsync for backups based on the rsync command using SSH protocol.

Default: rsync

Note: When backing up standby nodes with backup_method = rsync the minimum supported version of Barman is 2.8.

Dependencies

This role does not have any dependencies, but packages repositories should have been configured beforehand with the setup_repo role.

Example Playbook

Inventory file content

Content of the inventory.yml file:

---
all:
  children:
    barmanserver:
      hosts:
        barman1:
          ansible_host: xxx.xxx.xxx.xxx
          private_ip: xxx.xxx.xxx.xxx
    primary:
      hosts:
        primary1:
          ansible_host: xxx.xxx.xxx.xxx
          private_ip: xxx.xxx.xxx.xxx
          barman: true
          barman_server_private_ip: xxx.xxx.xxx.xxx
          barman_backup_method: rsync
    standby:
      hosts:
        standby1:
          ansible_host: xxx.xxx.xxx.xxx
          private_ip: xxx.xxx.xxx.xxx
          upstream_node_private_ip: xxx.xxx.xxx.xxx
          replication_type: synchronous
          barman: true
          barman_server_private_ip: xxx.xxx.xxx.xxx
          barman_backup_method: postgres
        standby2:
          ansible_host: xxx.xxx.xxx.xxx
          private_ip: xxx.xxx.xxx.xxx
          upstream_node_private_ip: xxx.xxx.xxx.xxx
          replication_type: asynchronous
          barman: true
          barman_server_private_ip: xxx.xxx.xxx.xxx
          barman_backup_method: postgres

In the High-Availability context, when using a Virtual IP address for routing read/write traffic to the primary node, then it's possible to configure barman to use this VIP address as Postgres connection endpoint. This way, in case of timeline change (following a failover or switchover event), barman will be able to automatically fetch new WAL records coming from the new primary node, without requiring any manual operation.

This configuration is only possible when using the postgres backup method, which is based on backup and WAL archiving using streaming replication.

Above is an example of inventory file implementing this type of configuration:

---
all:
  children:
    barmanserver:
      hosts:
        barman:
          ansible_host: 192.168.122.2
          private_ip: 192.168.122.2
    primary:
      hosts:
        pg1:
          ansible_host: 192.168.122.3
          private_ip: 192.168.122.3
          # Enable barman
          barman: true
          # Private IP address of the barman server
          barman_server_private_ip: 192.168.122.2
          # Backup and WAL archiving using the streaming protocol
          barman_backup_method: postgres
          # VIP of the primary node. Assuming this VIP address is already
          # present on the system.
          barman_primary_vip: 192.168.122.222
    standby:
      hosts:
        pg2:
          ansible_host: 192.168.122.4
          private_ip: 192.168.122.4
          upstream_node_private_ip: 192.168.122.3
          # Enable barman
          barman: true
          # Private IP address of the barman server
          barman_server_private_ip: 192.168.122.2
          # Backup and WAL archiving using the streaming protocol
          barman_backup_method: postgres
          # VIP of the primary node. Assuming this VIP address is already
          # present on the system.
          barman_primary_vip: 192.168.122.222
          # We don't want to create barman configuration and backup for this
          # instance.
          barman_no_configuration: true

How to include the setup_barman role in your Playbook

Below is an example of how to include the setup_barman role:

---
- hosts: primary, standby
  name: Configure Barman backup on Postgres nodes
  become: yes
  gather_facts: yes

  collections:
    - edb_devops.edb_postgres

  pre_tasks:
    - name: Initialize the user defined variables
      set_fact:
        pg_version: 14
        pg_type: "PG"

  roles:
    - setup_barman
      when: "'setup_barman' in lookup('edb_devops.edb_postgres.supported_roles', wantlist=True)"

Defining and adding variables is done in the set_fact of the pre_tasks.

All the variables are available at:

Important notes for recovery

This role adds the recovery_options=get-wal to the barman configuration file, this implies that the server where a backup is restored is able to access barman with ssh. If no ssh access is possible to the Barman server from the new PostgreSQL server then it is required to execute the recovery with: --no-get-wal parameter, like:

$ barman recover --no-get-wal \
     --remote-ssh-command 'ssh postgresql@restore_server' \
     my_server_instance \
     latest /path/to/new/pgdata

get-wal allows to recover partial WAL file from the backup when streaming backup is used (i.e. the one being currently written by the backuped PostgreSQL server).

barman recover --no-get-wal requires recovery_options=get-wal to be set.

License

BSD

Author information

Author: