Re: vacuum analyze corrupts database

Michael Brusser <michael(at)synchronicity(dot)com> writes:
> I should've done already - I loaded this table into another
> instance of the database and I could not reproduce the problem!

Hm. The log looks like the crash is occurring during planning (since
you get a "parse tree" message but no "ProcessQuery"). I'm guessing
that some selectivity-estimation operation that tries to use the
pg_statistic entries generated by VACUUM ANALYZE is failing. Can't
narrow it down further than that though. Again, how about a debugger
stack trace? Also, the pg_stats rows for the table would be interesting.

> If you're willing to look farther I could tar/zip the db as well.

Waste of time, likely, since I don't have a Solaris system to install it
on.

regards, tom lane