[postgis-users] memory bug in postgis query
Martin Güther
mguether at gmail.com
Sun Oct 23 04:03:03 PDT 2011
Hi guys,
i recently asked the same question on the postgresql mailing list but they forwarded me to this mailing list.
I really hope you can help me :)
on a test maschine from my university i'm running into a weird problem (i attached detail info):
A running postgis query eats more and more memory.(seen in htop)
It'll start using virtual memory and after eating up that one, too, postgres will
crash with:
psql:/home/mguether/sql/benchmark.sql:6: NOTICE: std::bad_alloc
psql:/home/mguether/sql/benchmark.sql:6: ERROR: GEOS intersects()
threw an error!
once it gave me an
psql:/home/mguether/sql/benchmark.sql:6: ERROR: out of memory
DETAIL: Failed on request of size 182949.
after the crash, all memory/swap is free again, so it really is postgres :)
i don't really know how to debug it any more detailed or fix this
issue in any way.
Does anyone here have a hint, where the issue might be?
thanks,
martin
here some information:
the maschine:
16 GB RAM, i7 4cores x64 2.6.32-21-server ubuntu
installed postgres:
psql (PostgreSQL) 8.4.9
including postgis
POSTGIS="1.4.0" GEOS="3.1.0-CAPI-1.5.0" PROJ="Rel. 4.7.1, 23 September
2009" USE_STATS
i changed two parameters from the default config:
shared_buffers = 512MB
work_mem = 256MB
using osm2pgsql I important some GEO data, and want to execute this query:
SELECT a.osm_id, w.osm_id FROM "planet_osm_line" w,
"planet_osm_polygon" a WHERE ST_Intersects(w.way,a.way) AND w.highway
IS NOT NULL;
the EXPLAIN gives me:
Nested Loop (cost=0.00..26427424.16 rows=1511021 width=8)
Join Filter: _st_intersects(w.way, a.way)
-> Seq Scan on planet_osm_line w (cost=0.00..49213.10
rows=1137488 width=837)
Filter: (highway IS NOT NULL)
-> Index Scan using osm_areas_index on planet_osm_polygon a
(cost=0.00..23.10 rows=6 width=824)
Index Cond: (w.way && a.way)
(6 rows)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20111023/9f33e03e/attachment.html>
More information about the postgis-users
mailing list