"Mike Williams" <mike.williams@comodo.com> writes:
> The "target block" stays the same for extended periods, but does change.
Fascinating. Don't know what it means, but it's fascinating.
> Where does the 1663 number come from?
> <datadir/base>/743352/743420 is a real file, but I don't have any file 1663.
1663 is the OID of the pg_default tablespace; that's never going to
change unless you move the table to a non-default tablespace.
> The server is Gentoo, and I don't have an identical server to use. This
> machines HA partner is staying at 8.3.5 until we're completely and
> entirely sure 8.3.10 works right.
I wasn't really suggesting that you need an *identical* machine. If
this is a software problem, as seems moderately likely at this point,
it should be possible to reproduce it on a different machine. Even more
to the point, it's going to be difficult to find the problem without
tracing through the fault case with a debugger --- which isn't something
you're going to want to do on a production server. Have you got a test
machine you can load the data onto? Preferably one you could give a PG
developer access to?
<aside>Using gentoo for production servers is generally viewed with
suspicion around here; gentoo is a fine distro but, um, not especially
focused on stability. Still, there's no obvious reason to blame this on
the distro.</aside>
regards, tom lane
