chef cookbook for installing wal-e (postgres continuous archiving to s3)

• https://github.com/wal-e/wal-e
• http://www.postgresql.org/docs/9.1/static/continuous-archiving.html
• This recipe automates many of the steps from this excellent article: http://blog.opbeat.com/2013/01/07/postgresql-backup-to-s3-part-one/

See sample usage: here

This cookbook has been developed and tested on ubuntu servers only

• postgresql - optionally use databox-cookbook which includes postgresql
• python - wal-e is installed using python pip (package management system)
• git - required to install wal-e using pip over git

• ['wal_e']['aws_access_key_id'] : AWS Access Key
• ['wal_e']['aws_secret_access_key'] : AWS Secret
• ['wal_e']['wale_s3_prefix'] : S3 url for the bucket where postgres WAL and backups will be stored
  • i.e. s3://your-lower-case-bucket-name/whatever/wal-e/
• ['wal_e']['wale_git_install_reference'] : Specify which commit to install wal-e from
  • Default is HEAD, which will install from the HEAD of master
  • Optionally set this to a git commit SHA or a git TAG to install a specific version
    • see https://github.com/wal-e/wal-e/tags for stable versions of WAL-e
  • NOTE: the recipe uses a marker file with the value from this configuration to determine if it should install a different version of WAL-e, so re-running it when set to HEAD will only run the first time, it will not pull newer commits unless you update this to a value besides HEAD, i.e. a SHA or TAG

• ['wal_e']['aws_access_key_id'] : Same as wal_e::master, see above
• ['wal_e']['aws_secret_access_key'] : Same as wal_e::master, see above
• ['wal_e']['wale_s3_prefix'] : Same as wal_e::master, see above
• ['wal_e']['wale_git_install_reference'] Same as wal_e::master, see above
• ['wal_e']['recover'] : add recover options if desired
• ['wal_e']['recover']['recovery_target_timeline'] : default is 'latest'
  • see http://www.postgresql.org/docs/9.1/static/recovery-target-settings.html#RECOVERY-TARGET-TIMELINE
• ['wal_e']['recover']['recovery_target_time'] : default is nil
  • see http://www.postgresql.org/docs/9.1/static/recovery-target-settings.html#RECOVERY-TARGET-TIME

Include wal_e in your node's run_list:

{
  "name":"my_node",
  "run_list": [
    "recipe[wal_e::master]"
  ]
}

and specify all 3 of the S3 attributes needed by wal-e

"wal_e": {
  "aws_secret_access_key": "SECRET",
  "aws_access_key_id": "ACCESS_KEY",
  "wale_s3_prefix": "s3://your-lower-case-bucket-name/wal-e/or_whatever",
  "wale_git_install_reference": "v0.6.2"
}

Include wal_e in your node's run_list:

{
  "name":"my_node",
  "run_list": [
    "recipe[wal_e::recover]"
  ]
}

and specify all 3 of the S3 attributes needed by wal-e

optionally add the recover options

"wal_e": {
  "aws_secret_access_key": "SECRET",
  "aws_access_key_id": "ACCESS_KEY",
  "wale_s3_prefix": "s3://your-lower-case-bucket-name/wal-e/or_whatever",
  "wale_git_install_reference": "v0.6.2",
  "recover": {
    "recovery_target_timeline": "latest",
    "recovery_target_time": null
  }
}

WARNING: the recover recipe will delete all of the pg data directory before pulling WAL-e files from S3, you may want to recover to a new server instead of running recover on your master

• take the master offline
  • ensure that it will no longer write to S3 with WAL-e
• run the recover recipe to bring up a new server in recovery mode
• once the chef script has completed, you will want to manually verify the server is up and has recovered all data
  • ssh onto box and tail postgres log
    • sudo su - postgres
    • tail -f /var/log/postgresql/postgresql-9.1-main.log
  • once server is online verify data is present
    • psql -c "\l"
    • psql app1 -c "select * from sample_data_1;"
• Once you have verified the server is up and running and all data is present you will want to 'promote' this server to be the new master
  • change your chef configuration for this node
  • it is recommended to change the S3 endpoint to a new directory
    • see https://groups.google.com/d/msg/wal-e/Mh0zeczwBCw/xGZ7KSEvYiwJ for more information on maintaining separate prefixes for each server that becomes a primary (master) server
    • Example wale_s3_prefix
      • Original Master: s3://my-org/pg_cluster_1/wal-e-1
      • Time passes Original Master goes down, bring up Master2 and change the prefix
      • Master2: s3://my-org/pg_cluster_1/wal-e-2
      • Time passes and Master2 goes down, bring up Master3 and change the prefix
      • Master3: s3://my-org/pg_cluster_1/wal-e-3
  • provision it using chef again
  • and then verify that it is now writing WAL files to S3 on the new prefix directory

• add recipe for standby server
• move attributes to encrypted data bag
• allow configuration of cron settings for backup creation

• Fork the repository on Github
• Create a named feature branch (like add_component_x)
• Write your changes
• Submit a Pull Request using Github

wal-e cookbook is released under the MIT License.

See the LICENSE file

# if you want to continually install latest HEAD version of WAL-e 
# run this before the wal_e::common recipe executes - (don't do this in production)
file "/installs/wal-e-marker" do
  content " "
end

psql -c "show data_directory"
# /var/lib/postgresql/9.1/main

psql -c "show config_file"
# /etc/postgresql/9.1/main/postgresql.conf

/var/log/postgresql/postgresql-9.1-main.log 

sudo pip install -e git git://github.com/wal-e/wal-e.git@#egg=wal-e
sudo pip install -e git git://github.com/wal-e/[email protected]#egg=wal-e
sudo pip install -e git git://github.com/wal-e/wal-e.git@17417b9552d96881ce4a86cbe6bae24f8a5ef241#egg=wal-e

pip list
sudo pip uninstall --yes --quiet wal-e

/usr/local/bin/wal-e version
# => 0.7.2.dev