Thread: Fresh initdb contains a few deleted B-Tree pages

I would like to put on the record that there are a few deleted B-Tree
pages in a freshly initdb'd database. As my btreecheck tools shows:

mgd=# select bt_index_verify(indexrelid::regclass::text) from pg_index;
NOTICE:  00000: page 12 of index "pg_attribute_relid_attnam_index" is deleted
LOCATION:  bt_page_verify_worker, btreecheck.c:206
NOTICE:  00000: page 13 of index "pg_attribute_relid_attnam_index" is deleted
LOCATION:  bt_page_verify_worker, btreecheck.c:206
NOTICE:  00000: page 14 of index "pg_attribute_relid_attnam_index" is deleted
LOCATION:  bt_page_verify_worker, btreecheck.c:206
NOTICE:  00000: page 9 of index "pg_attribute_relid_attnum_index" is deleted
LOCATION:  bt_page_verify_worker, btreecheck.c:206

(contrib/pageinspect's bt_page_items() will show a similar message if
this is done manually)

This is not a new-to-9.4 issue. I am not suggesting that it's a real
problem that requires immediate fixing, but it is suboptimal. We may
wish to fix this someday.

-- 
Peter Geoghegan

<p dir="ltr">Why is this even a small concern? Initdb runs the bootstrap SQL which does various SQL operations so it's
notsurprising there are some updates creating garbage. Iirc we don't even rely on template0 being frozen any more.<p
dir="ltr">--<br /> greg 

On Tue, Jul 1, 2014 at 6:02 AM, Greg Stark <stark@mit.edu> wrote:
> Why is this even a small concern? Initdb runs the bootstrap SQL which does
> various SQL operations so it's not surprising there are some updates
> creating garbage. Iirc we don't even rely on template0 being frozen any
> more.

It's not surprising that some initdb updates create garbage, but the
extent to which they do did slightly surprise me.


-- 
Peter Geoghegan