Re: vacuum_cost_delay & VACUUM holding locks on GIST indexes

Ron Mayer <rm_pg@cheapcomplexdevices.com> writes:
> When you VACUUM a table with postgis indexes (perhaps GIST indexes
> in general?) it seems a lock is held on the table.

GIST isn't concurrent-safe.

> Any clever workarounds?

Fix GIST ;-)

You could shorten the intervals for which the lock is held by reducing
vacuum_mem, but this might be counterproductive overall.

            regards, tom lane