BTW, I was thinking that the DROP INDEX CONCURRENTLY logic needed to be:
1. Unset indisvalid, commit, wait out all reading transactions.
2. Unset indisready, commit, wait out all writing transactions.
3. Unset indislive, commit (with parent table relcache flush),
wait out all reading-or-writing transactions.
4. Drop the index.
However, I wonder whether we couldn't combine steps 2 and 3, ie once
there are no readers of the index go directly to the "dead" state.
I don't see a need for a period where the index isn't being inserted
into but is still used for HOT-safety decisions.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
