Re: BUG #15636: PostgreSQL 11.1 pg_basebackup backup to a CIFSdestination throws fsync error at end of backup - Mailing list pgsql-bugs

From John Klann
Subject Re: BUG #15636: PostgreSQL 11.1 pg_basebackup backup to a CIFSdestination throws fsync error at end of backup
Date
Msg-id CAHyX5+VGCkWb_81_v_JsqJKzOEbBqDzjX8JmNWA5ogJTXmXgWw@mail.gmail.com
Whole thread Raw
In response to Re: BUG #15636: PostgreSQL 11.1 pg_basebackup backup to a CIFSdestination throws fsync error at end of backup  (Stephen Frost <sfrost@snowman.net>)
Responses Re: BUG #15636: PostgreSQL 11.1 pg_basebackup backup to a CIFSdestination throws fsync error at end of backup  (Thomas Munro <thomas.munro@gmail.com>)
List pgsql-bugs
Thanks Stephen for pointing this out.

John

On Mon, Feb 18, 2019 at 11:26 AM Stephen Frost <sfrost@snowman.net> wrote:
Greetings,

* PG Bug reporting form (noreply@postgresql.org) wrote:
>                               archive_mode = 'on'
>                               archive_command = 'test ! -f /pgxlog1/data1/%f || cp /pgxlog1/data1/%f
> /cifs/backups/<backupDirectoryName>/dmp/archive/%f'

I realize you were complaining about pg_basebackup here, but since you
shared your config, I wanted to point out that the above is an
absolutely *terrible* archive_command, *especially* when you've got a
network filesystem involved.

There's no guarantee with that archive command that the WAL makes it to
the backup server or that it's completely written out on the backup
server before the PG server removes the WAL- if something happened (a
network hiccup, the CIFS server being rebooted, or the PG server
restarting) there's a real possibility that you'd lose some portion of
your WAL, which could make a backup invalid, or could make it so you
can't do PITR.

Please don't roll your own backup solution like this.

Thanks!

Stephen

pgsql-bugs by date:

Previous
From: David Rowley
Date:
Subject: Re: BUG #15572: Misleading message reported by "Drop functionoperation" on DB with functions having same name
Next
From: Julien Rouhaud
Date:
Subject: Re: BUG #15572: Misleading message reported by "Drop functionoperation" on DB with functions having same name