Re: lazy vacuum sleeps with exclusive lock on table

[ITAGAKI Takahiro <itagaki(dot)takahiro(at)oss(dot)ntt(dot)co(dot)jp>]

Alvaro Herrera <alvherre(at)commandprompt(dot)com> wrote:

> What I'm requesting here is that the sleep in count_nondeletable_pages()
> be removed and that change backpatched to 8.2 and 8.1.

Agreed. We'd better to shorten the exclusive locking as far as possible.

> We don't know how many pages we can truncate until after we have
> acquired the exclusive lock and examined the pages in question, scanning
> backwards from the end of the table.

But many OSes do not care about read-ahead in backward scanning. I have
a test result that shows truncating a large part of table takes very long
time. Is it better better to change it to forward scanning? For example,
starting with (tail of the file - 16MB) and scanning 16MB of segment forward
to decide the position for truncation. If we can truncate all of the segment,
do recheck from (tail of the file - 32MB) and repeat.

Regards,
---
ITAGAKI Takahiro
NTT Open Source Software Center