Am 22.02.2017 um 16:00 schrieb Adrian Klaver:
On 02/22/2017 04:51 AM, Thomas Güttler wrote:
I have other concerns: atomar transaction. Movement should happen
completely or not all.
I don't think you can do this reliable (atomic transaction) with
"copy table_name".
You can if you wrap it in a transaction:
I want to **move** the data. The data should get deleted on the
satellite after transfer.
Well the replication suggestion is out.
I don't know how to delete the data which was copied, since inserts can
happen during the copy statement.
However you end up doing this I think you will probably need some sort of flag
on the rows on the satellites. It could
be a timestamp field of when the rows where inserted on the satellite or a
boolean field(copied). First instinct is to
use an insert timestamp and a tracking table that stores the last timestamp
used to move rows, where the timestamp is
only written on a successful transfer. To improve the chances of successful
transfer more smaller transfer batches
rather then larger transfers.
I really need a solid solution.
You said "... improve the chances of successful transfer ...". This makes me
nervous.
Delays are no problems, but data loss or duplication is.
Regards,
Thomas
--
Thomas Guettler http://www.thomas-guettler.de/
--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
