18/08/2017 17:45, Alvaro Herrera пишет:
> Erikjan Rijkers wrote:
>> On 2017-08-18 11:12, Ildar Musin wrote:
>>> Hi hackers,
>>>
>>> While we've been developing pg_pathman extension one of the most
>>> frequent questions we got from our users was about global index
>>> support. We cannot provide it within an extension. And I couldn't find
>>> any recent discussion about someone implementing it. So I'm thinking
>>> about giving it a shot and start working on a patch for postgres.
>> Sorry to be dense but what exactly is a "Global Index"?
> A global index covers all partitions of a partitioned table. It allows
> you to have unique indexes across the partitioned table.
>
> The main disadvantage of global indexes is that you need some kind of
> cleanup after you drop a partition. Either make partition drop wait
> until all the index pointers are removed, or you need some kind of
> after-commit cleanup process that removes them afterwards (which
> requires some assurance that they are really all gone). You can't let
> them linger forever, or you risk a new partition that reuses the same
> OID causing the whole index to become automatically corrupt.
>
Thanks for the notion, I haven't thought this through yet. We could
probably keep the list of removed partitions somewhere in the catalog or
in the index itself. Usually autovacuum (or the process we start right
after partition drop as you suggested) would clean the index up before
OID wraparound. But if it didn't and user is trying to add a new
partition with the same oid then the cleanup will be forced. I think the
latter is very unlikely.
--
Ildar Musin
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company