> I do think we need some better instrumentation for this kind of thing.
Well, one thing's for sure - I have little other information to offer. The
problem is that the lockups occur after hours of operation and thousands of
rows being digested (which is the nature of the program). If "better
instrumentation" implies tools to inpsect the sate of the db server's
process and to know what it's waiting for from the OS, I agree.
Then again, I can't even tell you whether the postgres process is at fault
or the TCL interface - which would be odd, because it's one fo the most
mature interfaces postgres has. So, here's a thought: is there any way for
me to inspect the state of a postgres process to see if it's responsive -
even if it's serving another connection?
Carlo
