Earlier I imported europe-latest.osm.pbf to database.
Yesterday I startet to pre-generate tiles from a portion of the europe area as set by this bounding box:
Now generate_tiles_multiprocess.py has only generated 275 tiles in 18 hours and 15 minutes.
I have set generate_tiles_multiprocess.py to use 4 threads:
NUM_THREADS = 4
as well as specifying this at the end:
minZoom = 1
maxZoom = 11
bbox = (3.98,57.69,32.23,71.46)
render_tiles(bbox, mapfile, tile_dir, minZoom, maxZoom, “Norway”)
I can see that the four CPU cores has a load usage of approximately 10 - 30%, most of the times it is about 20%.
More than 10GB of memory is not in use.
I do not understand why it is so slow.
But I do remember that I forgot to remove -l from the options when importing data to the database using osm2pgsql:
I used this:
osm2pgsql -s -U postgres -C 8000 --number-processes 4 -l -m -d osm -p planet_osm -E 3857 -S “N:\osm2pgsql\d
Can that -l option (–latlng) when importing data using osm2pgsql - cause Mapnik to use extra long time to generate the tiles?
If that option is the cause - can I fix this without having to import whole pbf file from Europe again?
If it’s not the cause, what could be the reason for the extremely slow tile generation?
- The whole DB on an SSD
- Tiledir on a 10000 RPM disk (SATA 2).
- 3.2 Ghz CPU with 4 cores.
- 20 GB RAM (5 x 4 GB)
- Windows 7