Re: Problem with VACUUM after very large delete?

At 06:12 PM 12/17/2002, Robert Treat wrote:
>While vacuum will delete dead index entries, it doesn't collapse out
>empty pages in the indexes. This causes indexes to never get smaller
>and, in cases where the range of the index continually expands, causes
>indexes to grow larger and larger. The generally accepted way to fix
>this is via reindexing. (note this was not fixed in 7.3)

Thanks Robert. However, if I can quote from my previous message:

pexicast_lg=# VACUUM VERBOSE audit_log;
NOTICE: --Relation audit_log--
NOTICE: Pages 2083605: Changed 2961, Empty 0; Tup 105773248: Vac 0, Keep
31106432, UnUsed 2.
Total CPU 89.49s/13.67u sec elapsed 1101.17 sec.
VACUUM

You'll notice that there are no "deleted" tuples listed. However, there are
"Keep" tuples listed. I'm not sure what those are - but that's almost
exactly the size of what I attempted to delete earlier. What are these
"Keep" tuples?

Also - whenever I vacuum my other tables, it does vacuum the indexes, with
additional NOTICE statements. In this case, it does not.

What's up?

Thanks,

Doug