[Tile-serving] [openstreetmap/osm2pgsql] Problem with query to get changed parent objects (Issue #2056)

[Sarah Hoffmann]

Sorry if that wasn't clear but the underlying issue is indeed the bad statistics for GIN indexes on array columns. That's a well known and long-standing issue. Last time it did bite us was when JIT and parallel indexing was introduced (#1045).

Problem is that our tables have now reached a size where Postgres resorts to sequential scans instead of using the index. This is bad because forcing Postgres to use an index is much more difficult than preventing it from using an index or JIT or parallel processing. The ideas above are about how to best do that without breaking things in a slightly different setup.

Tried `CREATE STATISTICS` but it made no difference. Maybe I did something wrong because it only took a few seconds to create it. I would have expected to take longer.

`enable_material = off` helps and is the workaround I'm using currently to get updates through on that machine. It looks like it slows down other queries of the update process. Updates are still very slow but at least not at a stand-still. So not a permanent solution.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/issues/2056#issuecomment-1694332637
You are receiving this because you are subscribed to this thread.

Message ID: <openstreetmap/osm2pgsql/issues/2056/1694332637 at github.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20230826/77214e82/attachment.htm>