cerxes/openmaptiles
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

[WIP] NUTS layer:

Browse Source 
Municipalities, Provinces, Regions etc as linestrings (with polygons used in intermediate step) representing what is left/right of the border
This commit is contained in:
Miel Truyen
2021-06-16 21:12:35 +02:00
parent 2a39e1ebfe
commit 7a239b66cf
10 changed files with 882 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
.env
View File

@@ -16,11 +16,13 @@ PGPORT=5432
# BBOX may get overwritten by the computed bbox of the specific area:
# make generate-bbox-file
# By default, the Makefile will use the content of data/$(area).bbox file if it exists.
BBOX=-180.0,-85.0511,180.0,85.0511
#BBOX=-180.0,-85.0511,180.0,85.0511
BBOX=4.827919,50.740472,5.798035,51.095291
# Which zooms to generate in make generate-tiles
MIN_ZOOM=0
MAX_ZOOM=7
MAX_ZOOM=14
# Use true (case sensitive) to allow data updates
DIFF_MODE=false
@@ -44,10 +46,28 @@ BORDERS_PBF_FILE=data/borders/filtered.pbf
BORDERS_CSV_FILE=data/borders/lines.csv
# Number of parallel processes to use when importing sql files
MAX_PARALLEL_PSQL=5
#MAX_PARALLEL_PSQL=5
MAX_PARALLEL_PSQL=16
# Number of parallel threads to use when generating vector map tiles
COPY_CONCURRENCY=10
# Number of concurrent IO ops to use when generating vector map tiles
# Set to ridiculous high values for SSD/NVME
# Test results: (These were without setting UV_THREADPOOL_SIZE)
# 64: ~100/s peak
# 256: ~200/s peak, stabilizes to a 50% improvement over 64 concurrency
# 1024: ~400/s peak, still at 400/s speed after a few minutes (unlike before, oddly enough)
# 4096: ~250/s peak, stabilizes to 200. But it appears to be choking on something, only updates progress periodically
# 1024 & 24 Threads: ~700/s peak - drops to 214s after 10mins, 178/s after 15mins. EST after 20min: 23hours
# Postgres is mainly idle during all of this (coarse dataset), using only half a core
# To be tested if 1 core is dedicated to postgres and the others to tilelive(UV_THREADPOOL_SIZE)
# will prove to be more efficient on average. Or just UV_THREADPOOL_SIZE==CPU_THREADS.
#COPY_CONCURRENCY=64
#COPY_CONCURRENCY=256
#COPY_CONCURRENCY=512
COPY_CONCURRENCY=1024
#COPY_CONCURRENCY=4096
UV_THREADPOOL_SIZE=16
#UV_THREADPOOL_SIZE=24
# Variables for generate tiles using PGquery
PGHOSTS_LIST=