MapComplete/Docs/SettingUpPSQL.md

# Setting up a synced OSM-server for quick layer access

## Requirements

Debian machine (e.g. Ubuntu)

`apt install osmium-tool docker.io osm2pgsql`

## Tuning the database



## Setting up the SQL-server (only once):


For performance, docker needs a specific config file. See https://osm2pgsql.org/doc/manual.html#tuning-the-postgresql-server

A suggested config file is:

(On HDD, 180GB RAM)

```
max_connections = 5000 # needed for OSM2PGSQL
shared_buffers = 2GB
work_mem = 200MB
maintenance_work_mem = 15GB
autovacuum_work_mem = 4GB
wal_level = minimal
checkpoint_timeout = 120min
max_wal_size = 20GB
checkpoint_completion_target = 0.9
max_wal_senders = 0
random_page_cost = 8.0
```
Save this config file somewhere as "postgresql.conf"

- Start the docker with: `docker run --name osm-cache-db -e POSTGRES_PASSWORD=password -e POSTGRES_USER=user -d -p 5444:5432 -v ~/data/pgsql/:/var/lib/postgresql/data -v postgresql.conf:/var/lib/postgresql/data/postgresql.conf postgis/postgis`
- Now, change the config file. This can easily be done as the config file is mirrored to the host system at:
`~/data/pgsql/postgresql.conf`. Append this file with configuration above, e.g. with `cat posgresql.conf >> ~/data/pgsql/postgresql.conf`
- Restart the db: `docker restart osm-cache-db`

# Seeding the data

> The following steps are also automated in .forgejo/workflows/update_database.yml

## Importing data

Install osm2pgsql (hint: compile from source is painless)

Download the latest with:
`nohup transmission-cli https://planet.osm.org/pbf/planet-latest.osm.pbf.torrent &>nohup_transmission.log &`
which will download the data to `~/Downloads`

To seed the database:
````
nohup osm2pgsql -O flex -S build_db.lua -s --flat-nodes=import-help-file -d postgresql://user:password@localhost:5444/osm-poi <file>.osm.pbf >> seeddb.log

# To see the progress
tail -f seeddb.log
````
Storing properties to table '"public"."osm2pgsql_properties" takes about 25 minutes with planet.osm

Belgium (~555mb) takes 15m
World (80GB) should take 15m*160 = 2400m = 40hr

73G Jan 23 00:22 planet-240115.osm.pbf: 2024-02-10 16:45:11  osm2pgsql took 871615s (242h 6m 55s; 10 days) overall on lain.local with RAID5 on 4 HDD disks, database is over 1Terrabyte (!)

Server specs

Lenovo thinkserver RD350, Intel Xeon E5-2600, 2Rx4 PC3 
    11 watt powered off, 73 watt idle, ~100 watt when importing

HP ProLiant DL360 G7 (1U): 2Rx4 DDR3-memory (PC3)
    Intel Xeon X56**


## Updating data


`osm2pgsql-replication update -d postgresql://user:password@localhost:5444/osm-poi -- -O flex -S build_db.lua -s --flat-nodes=import-help-file`


## Deploying a tile server

pg_tileserv can be downloaded here: https://github.com/CrunchyData/pg_tileserv

In the directory where it is downloaded (e.g. `~/data`), run

First, look up the latest suitable database on https://mapcomplete.org/status

````
export DATABASE_URL=postgresql://user:password@localhost:5444/osm-poi.{date-of-suitable-database}
nohup ./pg_tileserv >> pg_tileserv.log &
````

Tiles are available at: 
````
map.addSource("drinking_water", {
"type": "vector",
"tiles": ["http://127.0.0.2:7800/public.drinking_water/{z}/{x}/{y}.pbf"] // http://127.0.0.2:7800/public.drinking_water.json",
})
````

# Starting the summary server

`npm run summary-server` in the git repo

# Connecting to the database

Setup the SSH-tunnel:

`ssh folky.me
ssh -L 5444:127.0.0.1:5444 lain`

Open a second terminal:

`ssh -L 5444:127.0.0.1:5444 folky.me` `

Connect pgAdmin

# Rebooting:

-> Restart the docker container