On Tue, Jun 29, 2021 at 07:36:47PM +0100, Simon Riggs wrote: > Perhaps we should also add this text from the commit message to ensure > the importance is understood: > "This is extremely useful in cases where CIC/RC can run for a very long > time, because that used to be a significant headache for concurrent > vacuuming of other tables."
Uh, hopefully this idea is captured below. > Proposed edits: > > * "during certain index operations" -> "while concurrent index > operations run on other tables" Uh, the problem here is that it isn't clear if we are talking about concurrent index operations, "concurrently"-enabled index options, or "concurrently"-enabled concurrent index operations. I went with "minimal-locking index operations". > * spell Alvaro's name correctly Uh, this is the second time Mutt's default to using iso-8859-1 for non-ASCII attachments has caused confusion. I found that using 'set send_charset="us-ascii:utf-8"' in Mutt fixed this, so the attachment should now be fine. > * "row expiration" is a term not currently used in PG docs, so we > should probably look for something else. Yeah, I changed that to "removing dead rows" which seems to be our standard text. > There are 2 important features here, so the 2nd feature is worth > mentioning also: > > Avoid spurious waits in concurrent indexing > > Previously, multiple concurrent index operations could deadlock or > cause long waits. > Waits are avoided except for indexes with expressions, or WHERE predicates. OK, I added text to the bottom to try and capture that; new patch attached, now with UTF8 encoding. -- Bruce Momjian <[email protected]> https://momjian.us EDB https://enterprisedb.com If only the physical world exists, free will is an illusion.
diff --git a/doc/src/sgml/release-14.sgml b/doc/src/sgml/release-14.sgml index ebfe76b8cd..1556165533 100644 --- a/doc/src/sgml/release-14.sgml +++ b/doc/src/sgml/release-14.sgml @@ -742,6 +742,30 @@ Author: Peter Geoghegan <[email protected]> <listitem> <!-- +Author: Alvaro Herrera <[email protected]> +2020-11-25 [c98763bf5] Avoid spurious waits in concurrent indexing +Author: Alvaro Herrera <[email protected]> +2021-01-15 [f9900df5f] Avoid spurious wait in concurrent reindex +Author: Alvaro Herrera <[email protected]> +2021-02-23 [d9d076222] VACUUM: ignore indexing operations with CONCURRENTLY +--> + + <para> + Allow vacuum to be more aggressive in removing dead rows during + minimal-locking index operations (Álvaro Herrera) + </para> + + <para> + Specifically, <command>CREATE INDEX CONCURRENTLY</command> and + <command>REINDEX CONCURRENTLY</command> no longer limit the dead + row removal of other relations. This also avoids index commands + using <literal>CONCURRENTLY</command> from waiting for completion + of other operations using <literal>CONCURRENTLY</command>. + </para> + </listitem> + + <listitem> +<!-- Author: Fujii Masao <[email protected]> 2020-12-08 [e2ac3fed3] Speed up rechecking if relation needs to be vacuumed or -->
