old, dead connections?

Hi all,

I'm looking to upgrade a server to postgres 8.3, but there's one
connection that's being problematic. It continually says it's
UPDATE'ing, but it's been at it for weeks now, and while the database
is big, it shouldn't take weeks.

Even weirder is that it looks like the client-side connection died a
long time ago.

[root(at)haddock] ~ $ ps axvw | grep post
330 pts/0 S+ 0:00 0 75 3820 648 0.0 grep post
14321 ? S 0:00 9 3274 1056081 20428 0.6
/s/postgresql-8.2.6/bin/postgres -D /scratch.1/postgres/fsstudy-8.2
14323 ? Ss 19:31 16 3274 1055873 1036904 30.7 postgres:
writer process
14324 ? Ss 0:00 0 3274 8257 1644 0.0 postgres:
stats collector process
20459 pts/8 S+ 0:01 3 227 5480 3444 0.1
/s/postgresql/bin/psql -h haddock -p 5432 fsstudy
20460 ? Ss 2295:51 27 3274 1057221 1045304 31.0 postgres:
nitina fsstudy haddock.cs.wisc.edu(50957) idle
32247 ? Rs 17051:22 1 3274 1056653 1043964 30.9 postgres:
nitina fsstudy haddock.cs.wisc.edu(41114) UPDATE

[root(at)haddock] ~ $ netstat
Active Internet connections (w/o servers)
Proto Recv-Q Send-Q Local Address Foreign Address
State
tcp 0 0 haddock.cs.wisc.:postgresql
haddock.cs.wisc.edu:50957 ESTABLISHED
tcp 0 208 haddock.cs.wisc.edu:ssh ator.cs.wisc.edu:46929
ESTABLISHED
tcp 0 0 haddock.cs.wisc.edu:50957
haddock.cs.wisc.:postgresql ESTABLISHED
udp 0 0 localhost:33544 localhost:33544
ESTABLISHED
...

I'm fairly certain that the connections noted in netstat refer to the
idle connection.

So, I don't want to just kill the connection right now unless I'm
certain it can be done safely. However, I'd like to deal with it as
it's a potential impediment to the upgrade and it's taking up lots of
CPU time on the machine.

Any advice, or any idea of what's going on?

Peter