Skip to content

Latest commit

 

History

History
33 lines (19 loc) · 1.96 KB

pitr_wal-g.md

File metadata and controls

33 lines (19 loc) · 1.96 KB
Raw

Point in time recovery with wal-g

wal-g is the successor of wal-e, which no longer seems to be in active development! This example shows how to do point in time recovery with stolon using wal-g

wal-g correctly suggests to not put environment variables containing secret data (like aws secret keys) inside the archive_command since every user connected to postgres could read them. In its examples wal-g suggests to use the envdir command to set the wal-g required environment variables or (since some distribution don't have it) just use a custom script that sets them.

Backups

Base backups

Take the base backups using the wal-g backup-push command.

Archive backups

For doing this you should set at least the archive_mode and the archive_command pgParameters in the cluster spec. Wal-g will be used as the archive command:

stolonctl update --patch '{ "pgParameters" : { "archive_mode": "on", "archive_command": "envdir /etc/wal-g.d/env wal-g wal-push %p" } }'

Execute a point in time recovery

Note: looks like wal-g doesn't backups various config files like postgresql.conf, pg_hba.conf. While pg_hba.conf is currently generated by stolon, you'd like to keep the previous postgres parameters after the restore. For doing this there're two different ways:

  • if you want to backup the postgresql.conf you should do this outside wal-g. To restore it you have to create a dataRestoreCommand that will restore it after the wal-g backup fetch command.
  • if you don't want to backup/restore it than you can just set all the pgParameters inside the cluster specification
stolonctl init '{ "initMode": "pitr", "pitrConfig": { "dataRestoreCommand": "envdir /etc/wal-g.d/env wal-g backup-fetch %d LATEST" , "archiveRecoverySettings": { "restoreCommand": "envdir /etc/wal-g.d/env wal-g wal-fetch \"%f\" \"%p\"" } } }'