The scenario I'm looking at...
* Every night the database is able to be taken down for full backup ..
* During the runtime, the WAL files would be copied to an NFS mount
archivedir
* Let's say I keep a week of daily full backups and a week of archived
WAL files
-- Over time, there may be the case where archive dir could have a
duplicate name due to WAL name reuse
-- to avoid this, putting a timestamp on the filename in archivedir
would allow the archive WALS to accumulate without overwrite
I propose this scenario due to a past experience of table-level data
corruption that was not noticed by the Subject Matter Experts for over a
week and it was only by having multiple weeks worth of backups &
redo-logs that the database was able to be corrected...
Would a basic restore_command (restore_command = 'cp
/mnt/server/archivedir/%f %p' ) care if the ../archivedir/%f filenames
are "00000001000000010000000D.2012-0411.021245" (.YYYY-MMDD.HHMMSS) vs.
just "00000001000000010000000D"
I'm assuming the restore process works with the files at hand, sorted by
file create/modify timestamp, and not explicit filename...
Thanx!
Jeff
Jeff L'Huillier ( La-Hill-Yer )
JEFF.LHUILLIER@onstar.com
....Snappy little phrases here just get me in trouble ....
-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Wednesday, April 11, 2012 10:09 AM
To: L'Huillier, Jeff
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] 9.0.4/7 WAL Archiving: archive_command add
date-time to cp of filename
"L'Huillier, Jeff" <Jeff.LHuillier@ONSTAR.com> writes:
> When enabling WAL archiving and setting up the archive_command, is it
> possible to add the date & time as an extension to the %f copied to
the
> archive directory in order to avoid overwriting a file of the same
name?
Why would you think that's necessary? Please see the recommendations
for archive scripts, which tell you to simply refuse to copy in such a
case.
regards, tom lane
