BUG #6068: automatic analyze runs endlessly

"Jeff Janes" <jeff.janes@gmail.com> writes:

Starting with commit b4b6923e03f4d29636a94f6f4cc2f5cf6298b8c8,
"Fix VACUUM so that it always updates pg_class.reltuples/relpages."

After running make installcheck, the tables regression.public.slow_emp4000
and regression.public.fast_emp4000 get analyzed once a minute even though
they have no activity. None of the other installcheck tables, just those
two.

This has no direct negative effect that I know of, but it seems to be a
worrying symptom that pg_class or some related entity may not be getting
updated correctly.

Hmm ... seems worrisome, all right, but I don't see that happening here.
At least I don't see any indication of it in
pg_stat_all_tables.last_autoanalyze. Are you looking at that, or some
other evidence? Do you have any nondefault settings?

regards, tom lane

"Jeff Janes" <jeff.janes@gmail.com> writes:

After running make installcheck, the tables regression.public.slow_emp4000
and regression.public.fast_emp4000 get analyzed once a minute even though
they have no activity. None of the other installcheck tables, just those
two.

Oh, wait, I see what's different about those two tables: they have no
analyzeable columns. I thinko'd what to do in that situation. Will
fix.

regards, tom lane

On 6/19/11, Tom Lane <tgl@sss.pgh.pa.us> wrote:

"Jeff Janes" <jeff.janes@gmail.com> writes:

Starting with commit b4b6923e03f4d29636a94f6f4cc2f5cf6298b8c8,
"Fix VACUUM so that it always updates pg_class.reltuples/relpages."

After running make installcheck, the tables regression.public.slow_emp4000
and regression.public.fast_emp4000 get analyzed once a minute even though
they have no activity. None of the other installcheck tables, just those
two.

This has no direct negative effect that I know of, but it seems to be a
worrying symptom that pg_class or some related entity may not be getting
updated correctly.

Hmm ... seems worrisome, all right, but I don't see that happening here.
At least I don't see any indication of it in
pg_stat_all_tables.last_autoanalyze. Are you looking at that, or some
other evidence? Do you have any nondefault settings?

Just log_autovacuum_min_duration=0

Both evidence and setting.

Cheers, Jeff