Re: Automating backup
От | Neil Dugan |
---|---|
Тема | Re: Automating backup |
Дата | |
Msg-id | 43E7EDBB.8000706@butterflystitches.com.au обсуждение исходный текст |
Ответ на | Re: Automating backup (Richard Sydney-Smith <richard@ibisau.com>) |
Список | pgsql-general |
Richard Sydney-Smith wrote: > Hi Doug. > > Many users are haphazard in their approach until the machine fails and > then they expect to be pulled from the poo. > > Done it too many times. I now will get the application to enforce an > additional integrity check. It must be backed up or else! Seems futile > to put all the effort into a database design that checks and ensures > everything except that a backup copy exists! > > Essentially I need access to a database record that gives the timestamp > for the last backup. > Of course pgdump/vacuum could maintain such timestamps themselves. Is > that possible? If not then a proc in the database that calls > pgdump/vacuum and records the event is needed to give me peace of mind. > > regards > > Richard > I have just implemented a backup scheme on my Linux box. /etc/cron.daily/postgres su postgres /var/lib/pgsql/maintain /var/lib/pgsql/maintain vacuumdb --all --full pg_dumpall --clean | gzip > /var/lib/pgsql/backups/pg_`date --iso`.gz # remove any backups older than 30 days (this is a little dangerous) tmpwatch 720 /var/lib/pgsql/backups I created a directory #> mkdir /var/lib/pgsql/backups I suppose you could get the /var/lib/pgsql/maintain script to generate a sql statement in a temporary file (e.g. /tmp/setbackupdate.sql) then use psql to then update a record in your database with a command like (psql -c /tmp/setbackupdate.sql) Regards Neil. > Doug McNaught wrote: > >>Richard Sydney-Smith <richard@ibisau.com> writes: >> >> >> >>> pch := pchar('pg_dump -C -h '+host+' -U '+usr+' -p '+pswd+ ' -f >>>'+bckup_path+' '+dbase); >>> >>>to postgres. >>> >>>as the operator is obviously logged in how do I >>>(1) trap their user id >>>(2) Send the call to pg_dump without knowing their password? >>> >>>I expect this is a well worn route and am hoping not to have to >>>reinvent a wheel. >>> >>> >> >>I don't think it's "well-worn" at all--everyone I've ever heard of >>runs pg_dump from a cron script. >> >>Why not have a shell script run by the operator that runs pg_dump and >>then calls psql to insert the log record (assuming the dump succeeds)? >>Putting the logic inside of the database doesn't seem to buy you >>anything AFAICS. >> >>-Doug >> >>---------------------------(end of broadcast)--------------------------- >>TIP 4: Have you searched our list archives? >> >> http://archives.postgresql.org >> >> >> >> >>
В списке pgsql-general по дате отправления: