"Roberts, Adam" <Adam.Roberts@echostar.com> writes:
> The following is a link to the last post we had with this same problem:
> http://openacs.org/forums/message-view?message_id=185837
Given the comment about 500M pg_clog, I would venture that you aren't
doing adequate vacuuming and have therefore run into a transaction ID
wraparound failure. 500Mb clog corresponds to 2 billion transactions
which is right about where things would blow up if you'd not done the
necessary database-wide vacuums; and the fact that clog hasn't gotten
truncated is pretty strong evidence that you didn't.
Please read the administrator's guide:
http://www.postgresql.org/docs/7.4/static/maintenance.html
(adjust for your PG version as needed)
regards, tom lane