I have a local global OSM server up and it is running, but the rendering speeds vary drastically. An example:
DEBUG: DONE TILE english 9 216-223 144-151 in 32.009 seconds
DONE TILE english 16 74992-74999 11408-11415 in 304.705 seconds
I have run the indexes from here: OSM PostGIS Tuning and it seems to not have made much of a difference. I would say for Z16 the average render time is about 190 seconds. This is sitting on the following hardware:
- Intel i9-10850K (20 Core)
- 128GB DDR4
- OS is sitting on 1TB NVME
- OSM Database is sitting on 4TB NVME
Here are my postgres settings (this all sits on postgres 12) some taken from here: osm2pgsql manual:
-
max_connections = 300
-
shared_buffers = 16GB
-
work_mem = 10GB
-
maintenance_work_mem = 10GB
-
autovacuum = on
-
effective_io_concurrency = 300
-
wal_level = minimal
-
fsync = off
-
synchronous_commit = off
-
max_wal_size = 10GB
-
min_wal_size = 80MB
-
checkpoint_completion_target = 0.9
-
max_wal_senders = 0
-
random_page_cost = 1.1
-
effective_cache_size = 70GB
asked
04 Apr '22, 19:09
jaredalv
41●2●2●5
accept rate:
0%
Do you have a Postgres slow query log so you can see (assuming the bottleneck is Postgres) what's taking the longest time?