OmniPITR

TODO

Even while developing first version of the OmniPITR, we know there are things that could be improved. But first things first - we have to finish development of base functionality before starting work on new stuff.

Here are the already known about missing features:

  • Add a way to specify "primary" destination for wal segments.

    This would be useful if you have WAL-slave in the same network, and additional slave, on a network link that can be down. Problems with non-primary destinations wouldn't stop replication to primary destinations.

  • Make it possible to provide remote source in omnipitr-restore and omnipitr-backup-slave (remote, via http/ftp)

  • Make it possible to provide multiple sources for omnipitr-restore

    This is for migration purposes - for example, when you're switching your walarchive from compressed to not-compressed.

  • Add support for xz compression (similar to lzma)

  • Add support for using ZFS/LVM snapshops.

    This has the benefit of keeping required xlogs to a minimum for backups.

  • make omnipitr-backup-* scripts verify that they are being run as user running PostgreSQL

DONE

  • [1.2.0] Support sending to cloud storage such as Amazon S3 - it's done via --dst-pipe option.

  • [1.2.0] make omnipitr-restore don't care about walarchive writability if -r option is not given

  • [1.2.0] Add option to skip archiving of xlogs to omnipitr-backup-*

    This is assuming the user is keeping their own xlog archive and will limit the ability to restore a standalone database.

  • [1.2.0] Ability to use %r in omnipitr-restore instead of relying on pg_controldata

    While definitely good thing, we will still keep code to use pg_controldata as %r is not available in 8.2.

  • [1.1.0] Provide support for --help option

    Well, it can be helpful to avoid having to type another command to get to docs

  • [1.1.0] Add support for config file

    This has the benefit over command line options that it lets you change the options without Pg restart (in omnipitr-restore case at least).

  • [1.1.0] Add --version option to all programs.

  • [1.1.0] Make temp directories inside given temp-dir, based on pid and/or random values - to allow easy running (for example) 2 slaves on the same machine.

  • [0.7.0] Have omnipitr-backup-* look at .backup file to determine exactly what xlogs it needs to keep for backup integrity.

  • [0.6.0] Deliver to multiple destinations in parallel

    When delivering wal segments to slave server, and to backup destination - we can reduce the time it takes, by delivering it in parallel.

  • [0.4.0] Add specialized program to serve as archive_cleanup_command - with support for compressed walarchive, and pausing removal

The OmniPITR project is Copyright (c) 2009-2013 OmniTI. All rights reserved.