Re: pgsql: Skip full index scan during cleanup of B-tree indexes when possi - Mailing list pgsql-committers

From Alexander Korotkov
Subject Re: pgsql: Skip full index scan during cleanup of B-tree indexes when possi
Date
Msg-id CAPpHfdtOBG697zHRosd48x-OFsXa+vejiBy2Tgx_s4MgRQ1cpg@mail.gmail.com
Whole thread Raw
In response to Re: pgsql: Skip full index scan during cleanup of B-tree indexes whenpossi  (Teodor Sigaev <teodor@sigaev.ru>)
List pgsql-committers
On Thu, Apr 19, 2018 at 9:19 AM, Teodor Sigaev <teodor@sigaev.ru> wrote:
    Another thing that I noticed is that the metapage stores
    btm_last_cleanup_num_heap_tuples as a float4, even though
    xl_btree_metadata stores it as a double. I suggest that both places
    store it as float8, to be consistent. (It should not be double because
    we always avoid using anything other types with explicit typedef'd
    widths in WAL records.)


Good catch, thank you!  I also agree that both these fields should be of float8 type.
Please, find attached bt-vacuum-cleanup-float8-num-heap-tuples.patch fixing that.

Hm. patch changes on-disk representation of btree meta page. And there is no support to upgrade meta page since 857f9c36 commit (and should not, because that versions wasn't relesead), to prevent possible false positive bug reports I suggest to bump catversion. Objections?

I agree, catversion should be bumped for this patch.

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company 

pgsql-committers by date:

Previous
From: Teodor Sigaev
Date:
Subject: pgsql: Handle XLOG_BTREE_META_CLEANUP in btree_desc() andbtree_identif
Next
From: Teodor Sigaev
Date:
Subject: pgsql: Adjust _bt_insertonpg() comments