WAL reducing size

On Mon, Dec 4, 2017 at 8:50 AM, sql2pg <sekhar.career@gmail.com> wrote:

Is Postgres removes the committed transaction after archive. LIke in SQL
Server , which removes the committed transactions after taking Log backup.

If I understand correctly, yes - eventually. The docs explain the behavior
and the configuration variables that control it.

https://www.postgresql.org/docs/10/static/wal-configuration.html

David J.

how about uncommitted(open transactions) . if a segment has 1 committed and
2 uncommitted transactions then will it keep the segment instead deleting it
after checkpoint , since it has 2 uncommitted transactions

--
Sent from: http://www.postgresql-archive.org/PostgreSQL-general-f1843780.html

sql2pg wrote:

how about uncommitted(open transactions) . if a segment has 1 committed and
2 uncommitted transactions then will it keep the segment instead deleting it
after checkpoint , since it has 2 uncommitted transactions

There is no connection between transaction commit and checkpoint
removing WAL files. WAL files are removed after two checkpoints have
occured, regardless of transactions that might be open. The on-disk
effects of those transactions that occured before the earlier checkpoint
are already on disk anyway. The commit record might appear at whatever
point in the future, and the rows will not become visible to other
transactions until then, regardless of what happened to the WAL
segments.

--
�lvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

On Tue, Dec 5, 2017 at 6:12 AM, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:

sql2pg wrote:

how about uncommitted(open transactions) . if a segment has 1 committed and
2 uncommitted transactions then will it keep the segment instead deleting it
after checkpoint , since it has 2 uncommitted transactions

There is no connection between transaction commit and checkpoint
removing WAL files. WAL files are removed after two checkpoints have
occured, regardless of transactions that might be open. The on-disk
effects of those transactions that occured before the earlier checkpoint
are already on disk anyway. The commit record might appear at whatever
point in the future, and the rows will not become visible to other
transactions until then, regardless of what happened to the WAL
segments.

Something that perhaps does not concern you directly, but in Postgres
11 WAL segments get recycled after one completed checkpoint thanks to
commit 4b0d28d.
--
Michael