El mar, 27-01-2015 a las 08:36 +0100, Thomas Kellerer escribió:
> Martin Bruset Solberg schrieb am 26.01.2015 um 13:17:
> > I am faced with a scenario of backing up a database with one large
> > (partitioned) table, and several smaller tables. The large table is
> > so big that backing it up is proving problematic both in terms of
> > storage and time spent backing up and restoring. The backup of the
> > table itself is also reduntant, as the table data can be recreated by
> > importing from the source files. This data is not mission critical
> > for the application, so a speedy recovery is not urgent. The data
> > from the smaller tables, however, is critical. The combined size of
> > the rest of the database is not a problem neither in terms of storage
> > or backup/restore speed, so a logical backup excluding the large
> > table/schema works fine. But logical backup is not sufficiant, as I
> > need to be able to do PITR in case of disaster.
> >
> > So my question is, what are my options when I want to make a PITR
> > compatible backup, and need to exclude a table/schema/tablespace of
> > the database? Any suggestions welcome...
>
>
> What if you change that big table to UNLOGGED?
> As unlogged tables don't generate WAL they don't contribute to the PITR "restore time"
> (and they won't be restored if you have to do a PITR!)
However, they are not crash-safe: an unlogged table is automatically
truncated after a crash or unclean shutdown.
>
> It will be included in the necessary base backup though if I'm not mistaken.
>
> Thomas
>
>
>