Sebastien Boisvert <sebastienboisvert@xxxxxxxxx> writes: > [ COPY fails to dump a 138MB bytea column ] If you can't switch to a 64-bit build of Postgres, you might need to think about converting those byteas to large objects. It's expected for COPY to require memory space equal to several times the width of the row it's trying to dump, because it has to have multiple copies of the row data in order to do format conversion. In this case the required space seems to be going well past 1GB, so I wonder whether you are doing anything that exacerbates the memory requirement, for instance by forcing an encoding conversion to something other than the database's server_encoding. But having said that, it seems a bit odd that it is failing at what seems to be only around 2GB of memory demand. Even in a 32-bit build I'd expect there to be enough address space for that, especially in view of your very small (arguably too small) shared_buffers setting. Are you sure the postmaster is really being started with 'unlimited' ulimits? regards, tom lane -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
