cron job to remove old data from postgres on debian

Question

I need to create a cron that runs every night and remove some of the data (older ones) from my database. I am running postgres on debian. How do I do that? My server is tomcat6. Does anyone have a step by step instruction including script to do that?

Answer 1

in most cases, I prefer a shell script with an here-document. (shell-variables expand nicely in here-documents):

#!/bin/sh
PSQL=/local/postgres/bin/psql
SOME_VALUE=123

$PSQL my_database <<THE_END
DELETE FROM my_table
WHERE my_column <= $SOME_VALUE
    ;
THE_END

Answer 2

Several good ways.
You may want to do more than just one DELETE. Run ANALYZE tbl? Handle multiple tables?
I would wrap the work in a a plpgsql function:

CREATE OR REPLACE FUNCTION myschema.f_maintain()
  RETURNS text AS
$BODY$
DECLARE
    _msg    text := 'Report';
    _min_ts timestamp := now()::timestamp - interval '5y'; -- max. age 5 years
    _count  int;
BEGIN

DELETE FROM tbl
WHERE  my_ts < _min_ts
AND    <some condition>;

GET DIAGNOSTICS _count = ROW_COUNT;

IF _count > 0 THEN
    _msg := _msg || E'\nDeleted rows in tbl: ' || _count;
END IF;

ANALYZE tbl;

-- more ?

RAISE LOG '%', _msg;

END;
$BODY$
  LANGUAGE plpgsql;

... and call it from the shell like this:

psql -p 5432 event -c 'SELECT myschema.f_maintain()'

Schedule a cron job for the system user postgres:

crontab -e

To run it every day at 02:06, add a line like:

06 02 * * * psql -p 5432 event -c 'SELECT myschema.f_maintain()' >/dev/null

Answer 3

You can put a direct call to psql right in /etc/crontab. The call would follow this pattern:

psql -c 'DELETE FROM my_table WHERE ...'

You would need to supply the appropriate additional arguments to 'psql' to create the database connection. See the documentation for psql for more information about that.