On 29 March 2017 at 16:44, Craig Ringer <craig@2ndquadrant.com> wrote:
> * Split oldestCatalogXmin tracking into separate patch
Regarding this, Simon raised concerns about xlog volume here.
It's pretty negligible.
We only write a new record when a vacuum runs after catalog_xmin
advances on the slot with the currently-lowest catalog_xmin (or, if
vacuum doesn't run reasonably soon, when the bgworker next looks).
So at worst on a fairly slow moving system or one with a super high
vacuum rate we'll write one per commit. But in most cases we'll write
a lot fewer than that. When running t/006_logical_decoding.pl for
example:
$ ../../../src/bin/pg_waldump/pg_waldump
tmp_check/data_master_daPa/pgdata/pg_wal/000000010000000000000001 |
grep CATALOG
rmgr: Transaction len (rec/tot): 4/ 30, tx: 0, lsn:
0/01648D50, prev 0/01648D18, desc: CATALOG_XMIN catalog_xmin 555
rmgr: Transaction len (rec/tot): 4/ 30, tx: 0, lsn:
0/0164C840, prev 0/0164C378, desc: CATALOG_XMIN catalog_xmin 0
pg_waldump: FATAL: error in WAL record at 0/16BBF10: invalid record
length at 0/16BBF88: wanted 24, got 0
and of course, none at all unless you use logical decoding.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
