Re: pg_basebackup ignores the existing data directory permissions - Mailing list pgsql-hackers

From Haribabu Kommi
Subject Re: pg_basebackup ignores the existing data directory permissions
Date
Msg-id CAJrrPGeUbGADUz8O4gF1Nng2+zqSoNzWOGrx_NDRpPREXn0+yw@mail.gmail.com
Whole thread Raw
In response to Re: pg_basebackup ignores the existing data directory permissions  (Michael Paquier <michael@paquier.xyz>)
Responses Re: Re: pg_basebackup ignores the existing data directory permissions
List pgsql-hackers

On Tue, Mar 26, 2019 at 1:27 PM Michael Paquier <michael@paquier.xyz> wrote:
On Sun, Mar 24, 2019 at 10:30:47PM +1100, Haribabu Kommi wrote:
> With the above additional options, the pg_basebackup is able to control
> the access permissions of the backup files, but when it comes to tar mode
> all the files are sent from the server and stored as it is in backup, to
> support
> tar mode group access mode control, the BASE BACKUP protocol is
> enhanced with new option GROUP_MODE 'none' or GROUP_MODE 'group'
> to control the file permissions before they are sent to backup. Sending
> GROUP_MODE to the server depends on the -g option received to the
> pg_basebackup utility.


Thanks for the review.
 
Do we really want to extend the replication protocol to control that?

As the backup data is passed in tar format and if the pg_basebackup
is also storing it in tar format, i feel changing the permissions on tar
creation is easier than regenerating the received tar with different 
permissions at pg_basebackup side.

Other than tar format, changing only in pg_basebackup can support
independent group access permissions of the standby directory.

I am really questioning if we should keep this stuff isolated within
pg_basebackup or not.  At the same time, it may be confusing to have
BASE_BACKUP only use the permissions inherited from the data
directory, so some input from folks maintaining an external backup
tool is welcome.

That would be good to hear what other external backup tool authors
think of this change.

Regards,
Haribabu Kommi
Fujitsu Australia

pgsql-hackers by date:

Previous
From: "Jamison, Kirk"
Date:
Subject: RE: Timeout parameters
Next
From: Thomas Munro
Date:
Subject: Re: Refactoring the checkpointer's fsync request queue