>The only difference from the database point of view, is that PCs transactions, since they are much more complex, usually start with a ROLLBACK (arguable, it will be eventually removed) and for devices they usually just start with a COMMIT.
Correction ... and for devices they usually just start with a BEGIN TRANSACTION. Real thing is that when this happens, I'm several kilometers away, with no remote access (yet at least), and since it is a production system as soon as something like this happens they need to get it back up asap. I find it extremely difficult to be able to gather that data. Besides that I'm not familiar with the procedure of attaching to someone else's process in windows. It is a 8.4.2 postgres running in windows 2003 server. Im tempted in upgrading to 8.4.latest, since it should not require a full backup/restore and practically no server downtime. But i'm not sure about the real gain of that. On Thu, Sep 2, 2010 at 2:56 PM, Tom Lane <t...@sss.pgh.pa.us> wrote: > Eduardo Piombino <drak...@gmail.com> writes: > > Do you have any idea on what could be causing a SELECT NOW() to never > come > > back? > > That's truly bizarre. Can you attach to one of these stuck processes > with a debugger and get a stack trace? > > regards, tom lane >