Thursday, 6 September 2012

How to get an up-to-date OSM in PostGIS

How to build an OSM database that is up to date. 

Rather than doing a quarterly manual job of downloading a 14Gb+ file, decompressing it (250Gb+) and inserting it into a database it will be a lot easier to download daily (50Mb) OSM change files (.osc) and apply the changes to the existing database. 

The instructions below is a modified version of Martin van Exel's tutorial here - https://docs.google.com/document/pub?id=1paaYsOakgJEYP380R70s4SGYq8ME3ASl-mweVi1DlQ4 

# get osmosis
su cd /opt 
wget http://dev.openstreetmap.org/~bretth/osmosis-build/osmosis-latest.tgz 
tar zxvf osmosis-latest.tgz
rm -rf osmosis-latest.tgz
# in my case osmosis-0.41 

# now pop into PostgreSQL and build a suitable database & user (could be tidier)
adduser osm 
passwd osm 
# and some badly chosen password: osm, osm
su - postgres psql 
createdb osm
createlang plpgsql osm
CREATE USER osm WITH PASSWORD 'osm';
#CREATE DATABASE osm;
GRANT ALL PRIVILEGES ON DATABASE osm to osm;
GRANT ALL PRIVILEGES ON DATABASE osm to postgres;
\q
psql -d osm -U osm -f /usr/pgsql-9.1/share/contrib/postgis-2.0/postgis.sql 
psql -d osm -U osm f /usr/pgsql-9.1/share/contrib/postgis-2.0/spatial_ref_sys.sql
psql -U osm -d osm
CREATE EXTENSION hstore;
\q
psql -U osm -d osm -f /home/<user name>/osm/osmosis-0.41/script/pgsimple_schema_0.6.sql

# make some directories and get some data
su mkdir /tmp/osm 
cd tmp/osm 
mkdir planet
mkdir planet/replication

# now get some OSM data - either get the whole planet.. or get a sample from http://download.geofabrik.de/osm/ 
cd /tmp/osm/planet/
wget http://planet.openstreetmap.org/planet-latest.osm.bz2

# and thump into the database (using the VERY IMPORTANT --slim option, -C is how much memory we're setting aside for this: 4Gb in this case) /opt/osm2pgsql/osm2pgsql -S default.style --slim -d osm -C 4000 /tmp/osm/planet/planet-latest.osm.bz2 

# create the configuration.txt and download.lock 
osmosis --rrii workingDirectory=/tmp/osm/planet/replication 

# create your state.txt file, visit http://toolserver.org/~mazder/replicate-sequences/ and enter details. To check the datetime required examine the contents of the bz2 file. 

bunzip2 -c planet-latest.osm.bz2 | head

Now at this point I've been running osmosis to grab the change file and osm2pgsql to throw it at the database but you can apparently run osmosis on it's own;

# Osmosis on it's own version
/opt/osmosis-0.41/bin/osmosis --rri workingDirectory=/tmp/osm/planet/replication --sc --wpc user="osm" database="osm" password="osm"

# Or the two command route - grab a change file from the queue of necessary changes to perform
/opt/osmosis-0.41/bin/osmosis --rri workingDirectory=/tmp/osm/planet/replication --simplify-change --write-xml-change /tmp/osm/planet/replication/changes.osc.gz

# And append the changes to the database
/opt/osm2pgsql/osm2pgsql --append -S /opt/osm2pgsql/default.style -d osm -U postgres --slim /tmp/osm/planet/replication/changes.osc.gz 

# You can put either set of commands into a script and cron it up, here's the script for the osm2pgsql route

#!/bin/sh
n=`ps -ef | grep -v grep | grep /opt/osmosis-0.41/ | wc -l`
m=`ps -ef | grep -v grep | grep /opt/osm2pgsql/ | wc -l`
let i=n+m
if [ $i -gt 0 ]; then
echo osmosis or osm2pgsql running
else
echo not running
/opt/osmosis-0.41/bin/osmosis --rri workingDirectory=/tmp/osm/planet/replication --simplify-change --write-xml-change /tmp/osm/planet/replication/changes.osc.gz
/opt/osm2pgsql/osm2pgsql --append -S /opt/osm2pgsql/default.style -d osm -U postgres --slim /tmp/osm/planet/replication/changes.osc.gz 
fi

# Now you're probably running mod_tile and mapnik to render and serve up OSM tiles so we need a method to tell mod_tile to re-render all tiles that have updates. We can amend the osm2pgsql call with -e options to produce a list of tiles that need to be expired and thus re-rendered. [See OSM Tile Expire Methods].
# change the osm2pgsql command to read
/opt/osm2pgsql/osm2pgsql --append -S /opt/osm2pgsql/default.style -d osm -U postgres --slim /tmp/osm/planet/replication/changes.osc.gz -e15 -o expire.list

# We now expire all tiles listed in the expire.list. I haven't included the install and setup of Mapnik or mod_tile so i'm going to use some dummy locations for them. I'm expiring tiles from zoom level 6 and downwards
cat expire.list | /opt/mapnik/mod_tile/render_expired --map=osm --min-zoom=6 --touch-from=7 >/dev/null

# The OSM Tile Expire Methods guide also recommends a daily re-render of the lower zoom levels.

No comments: