Horton

Re: [DOC] Document concurrent index builds waiting on each other

David G. Johnston <david.g.johnston@gmail.com>

View thread View on postgr.es
From: David Johnston <david.g.johnston@gmail.com>
To: pgsql-hackers@lists.postgresql.org
Cc: James Coleman <jtc331@gmail.com>
Date: 2020-07-16T23:33:23Z
Lists: pgsql-hackers 
The following review has been posted through the commitfest application:
make installcheck-world:  not tested
Implements feature:       not tested
Spec compliant:           not tested
Documentation:            tested, passed

James,

I'm on board with the point of pointing out explicitly the "concurrent index builds on multiple tables at the same time will not return on any one table until all have completed", with back-patching.  I do not believe the new paragraph is necessary though.  I'd suggest trying to weave it into the existing paragraph ending "Even then, however, the index may not be immediately usable for queries: in the worst case, it cannot be used as long as transactions exist that predate the start of the index build."  Adding "Notably, " in front of the existing sentence fragment above and tacking it onto the end probably suffices.

I don't actually don't whether this is true behavior though.  Is it something our tests do, or could, demonstrate?

It is sorta weird to say "one will not return until all have completed, though, since usually people think return means completed".  That whole paragraph is a bit unclear for the inexperienced DBA, in particular marked ready to use but isn't usable.

That isn't really on this patch to fix though, and the clarity around concurrent CIC seems worthwhile to add, even if imprecise - IMO it doesn't make that whole section any less clear and points out what seems to be a unique dynamic.  IOW I would send the simple fix (inline, not a new paragraph) to a committer.  The bigger doc reworking or actual behavioral improvements shouldn't hold up such a simple improvement.

David J.

The new status of this patch is: Waiting on Author

Commits

  1. Call out vacuum considerations in create index docs

  2. Document concurrent indexes waiting on each other

  3. snapshot scalability: Don't compute global horizons while building snapshots.

  4. Remove PROC_IN_ANALYZE and derived flags

  5. Improve performance of get_actual_variable_range with recently-dead tuples.