Alexandre Leclerc <aleclerc@ipso.ca> wrote:
> The vacuum raised a "max_fsm_pages" of 142000 not enought and
> stopped.
That's probably just a warning that it wasn't able to track all the
dead space -- I would expect that. You're going to want to clean up
the bloat anyway. I would try a pg_dumpall at this point. If it
works, you might just be in good shape. I would try hard to keep
the old database when restoring it, and I recommend VACUUM ANALYZE
after restoring it. Then I would do some sanity checks to make sure
it looks like all the data is there and sound.
If it doesn't let you in with normal connections because other
databases are close to wrap-around, you'll have to vacuum those, but
that should be fast. However, before any other vacuums on that
cluster, I would try to copy the database off to some backup medium
(with the service stopped). If you have time, it's a very good idea
anyway.
I hope it all works out.
-Kevin
