Installation Manual for Martin?

[hadjiprocopis]

I am absolutely a beginner with Martin.

I am trying to relieve the burden on the OpenStreetMap public tile servers by deploying my own local installation of OSM data/tiles to be used for my own apps etc.

I found the installation manual provided e.g. at [https://maplibre.org/martin/quick-start.html] insufficient (for me) for getting martin to run locally (in linux headless server) from scratch, including the database setup, using openstreetmap tiles and behind nginx.

There are fragments for setting up things, like nginx but I got nowhere except that I have successfully installed martin from source with cargo following the instructions provided in the manual:
cargo install martin --locked
that worked great.

I now have martin installed including 2 executables.

But then I would like to set up the database, as a local postgres. How?

Then I would like to populate that database with OSM data/tiles. How?

Then I want to set up nginx. I see in the manual that this section is
quite detailed, thanks. I will try these instructions as soon as I have martin running with the db and the tiles data.

I am asking to point me to a step-by-step guide on how to install martin in a headless linux server with postgres and OSM tiles for a specific region, e.g. Monaco. Is there such a thing somewhere? Without docker please. Any help appreciated.

thank you for the great work you do on open-source projects

[CommanderStorm]

For loading data into postgis, the tool you want to look at is osm2pgsql.
I am not sure if this is the best course of action for you.

Choosing a postgres db for your basemap is not going to be great performance wise.
For your base map you want it pre-tiled in an mbtiles/pmtiles format.
Doing this at runtime is fairly expensive for the higher zoom levels.

Here is our guide on how to set up this part:

https://maplibre.org/martin/recipe-basemap-postgis.html

You can run planetiler to generate your planet every day, the basemap usually does not need to be that fresh.
You can then overlay the dynamic data you need to be very fresh.

Does this answer your question?

Btw:
Osmf is happy to provide those vector tiles and unless you are very heavy, that is very fine.
If not, something like https://stadiamaps.com is much more affordable (and better for a global user base) than running your own tileserver.
It is fun though and you can do a few optimisations/customisations that you could not do otherwise.

[hadjiprocopis]

Ok, I did not know that the database part is optional (if not suboptimal).

I will persist at the moment with my own server (vs commercial) just for the fun of it.

Good news : I now have Martin up and running. This is what I did. All info found in your manuals but perhaps in different sections:

1. cargo install martin --locked
   Now I have martin executable accessible from my terminal.
2. clone the planetiler from git:
   git clone --recurse-submodules https://github.com/onthegomap/planetiler.git
3. build planetiler from source, into the planetiler dir (just git clone'ed)
   mvn clean install or mvn clean install -DskipTests to skip tests for the impatient optimist.
4. If all goes well you will have the planetiler jar at:
   planetiler-dist/target/planetiler-dist-0.9.3-SNAPSHOT-with-deps.jar (from inside the git clone'd planetiler dir)
5. Here is how I use planetiler to get the tiles I needed for my locality (from inside the git clone'd planetiler dir):
   java -Xmx1g -jar "planetiler-dist/target/planetiler-dist-0.9.3-SNAPSHOT-with-deps.jar" --download --area=monaco . This creates a data subdir at the current dir and should contain an output.mbtiles file.
6. Now we are ready to run Martin with said tiles data from file without any database: martin output.mbtiles
7. It is serving data at port 3000 of localhost. Test it by getting the "catalog" wget localhost:3000/catalog -O catalog. It should fetch a JSON file saved as catalog.
8. Now, I want that to be running behind NGINX, and so I close outside access of port 3000 of my headless server and configure NGINX using the configuration pointed to by your manual (i think) here: https://github.com/maplibre/martin/blob/main/demo/frontend/nginx.conf.
9. If all goes well I have access to Martin from the outside of my server using wget 'https://myserver.com/tiles/catalog' or wget 'https://myserver.com/tiles/health' or wget 'https://myserver.com/tiles/_/metrics' (add a --no-check-certificate if you have a self-signed SSL certificate at your server). And also because Martin's default port (3000) is now blocked by the firewall, this should fail: wget 'https://myserver.com:3000/catalog , as that's how I want it to be.
10. I have created a systemd service to run Martin automatically on boot (note that I am not an expert on systemd service files so this may not be optimal or adviced):

[Unit]
Description = Run Martin OSM tiles web service on port 3000
Requires = network.target
After = network-online.target

# limit the restart attempts in 2 per hour
StartLimitIntervalSec=1h
StartLimitBurst=2

[Service]
Type = simple
Restart = no
SyslogIdentifier = martintilesserver
PIDFile = /var/lock/martin.pid
User = xyz
ExecStart = martin /opt/martin/planetiler-data/data/monaco.mbtiles
KillSignal = SIGQUIT
WorkingDirectory = /opt/martin

[Install]
WantedBy = multi-user.target

So, it ends well. Thank you, a.

[CommanderStorm]

To maybe get feedback on the docs:
Why do you prefer to not use docker?

[hadjiprocopis]

I am thinking that docker is wasting my server resources by running e.g. postgres inside a virtual machine container when I can do exactly the same, one level up on my machine, directly without running those docker commands. But I understand that this is the way forward for most people.