From: | Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk> |
---|---|
To: | Andres Freund <andres(at)anarazel(dot)de> |
Cc: | Pavan Deolasee <pavan(dot)deolasee(at)gmail(dot)com>, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Toast issues with OldestXmin going backwards |
Date: | 2018-04-21 23:35:41 |
Message-ID: | 878t9gcf9j.fsf@news-spur.riddles.org.uk |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
>>>>> "Andrew" == Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk> writes:
Andrew> There is still some overhead in the check, of course; I will
Andrew> see about doing some benchmarks.
Preliminary result suggests that the user-CPU cost of vacuum full
increases by ~16% when the entire table is recently-dead rows (with a
toasted column of ~10k length) compared to the same table when all rows
are live.
Since I doubt the practical value of vacuum full on a table which is
100% recently-dead, and that I would expect the live:recently-dead ratio
to not normally be much worse than 1:1 (making the overhead 8%) and more
likely up around 10:1 (making the overhead 1.5%), I think this is not an
issue. (When you have a lot of recently-dead rows is exactly the _wrong_
time to be doing a vacuum full, since it wouldn't save you any space and
you'd bloat the toast table as mentioned previously.)
--
Andrew (irc:RhodiumToad)
From | Date | Subject | |
---|---|---|---|
Next Message | Peter Geoghegan | 2018-04-22 01:02:12 | Re: Corrupted btree index on HEAD because of covering indexes |
Previous Message | Andrew Gierth | 2018-04-21 22:58:53 | Re: Toast issues with OldestXmin going backwards |