init grafana service

[conscious-puppet]

resolves #59

[shivaraj-bh]

You can use the http_get readiness probe instead of curl.

See example: https://github.com/Platonic-Systems/process-compose-flake/blob/b346fb30ad5fdee14e3edc3ba490eaa5db9ec0b7/example/flake.nix#L56-L59

[srid]

Also, perhaps make the whole ${grafanaConfig.server.protocol}://${grafanaConfig.server.domain}:${builtins.toString grafanaConfig.server.http_port}" a value of an internal option (grafanaUri?), so it can be re-used in other places, including by the user. It also makes the probe CLI here more readable.

[conscious-puppet]

@srid sure.

grafana has a config called root_url for this. i'll store the resulting value in this config

[srid]

What happens if extraConfg = { server.http_port = 8080; }? Would it not end up removing protocol and domain? Shouldn't we do a deep merge?

[conscious-puppet]

i was not sure about this. i'll test this out

[conscious-puppet]

i am using lib.recursiveUpdate now.

i have made this same error here. i'll create a PR to fix this

• services-flake/nix/clickhouse/default.nix

  Line 95 in 76beca3

  } // config.extraConfig

[srid]

If we don't expect the user to set this, you want to add readOnly = true; here.

[conscious-puppet]

grafana actually allows to give a root_url with subpath. in that case, user may set this value.

ref: https://grafana.com/docs/grafana/latest/setup-grafana/configure-grafana/#root_url

[shivaraj-bh]

Will the user simply not set extraConfig.server.root_url instead? This additional option might not be needed. I think the latest impl. of readiness_probe looks good as is

[shivaraj-bh]

Suggested change

	1. Create `postgres` service.
	```nix
	services.postgres."pg1" = {
	enable = true;
	listen_addresses = "127.0.0.1";
	port = 5435;
	initialDatabases = [{
	name = "grafana-db";
	}];
	initialScript.after = ''
	CREATE USER gfuser with PASSWORD 'gfpassword' SUPERUSER;
	'';
	};
	```
	2. Create `grafana` service, and change its config to use the `postgres` database.
	```nix
	services.grafana."gf1" = {
	enable = true;
	http_port = 3001;
	extraConf = {
	database = {
	type = "postgres";
	host = "127.0.0.1:5435";
	name = "grafana-db";
	user = "gfuser";
	password = "gfpassword";
	};
	};
	};
	```
	3. Add a setting to start `grafana` only after `postgres` is running.
	```nix
	settings.processes."gf1".depends_on."pg1".condition = "process_healthy";
	```
	```nix
	{
	services.postgres.pg1 = {
	enable = true;
	listen_addresses = “127.0.0.1”;
	};
	services.grafana.gf1 = {
	enable = true;
	extraConf.database = with config.services.postgres.pg1; {
	type = “postgres”;
	host = “${listen_addresses}:${port}”;
	};
	};
	settings.processes."gf1".depends_on."pg1".condition = "process_healthy";
	}
	```

It would be nice to keep these tips/suggestions simple and short.
I have assumed that not specifying name and user for postgres db in grafana will use postgres by default for both, if not we can explicitly mention.

[shivaraj-bh]

initialDatabases and intialScript can be part of postgres doc (whenever we add it) and if one wants to play around with postgres configs, they can refer from there.

I say this because, we are better off keeping these mini-tutorials short (with default configs) and not overwhelm the reader.

[conscious-puppet]

sure, by default grafana looks for the user root in database. and for the database grafana. i have added config these two over your suggestion. please have a look

[shivaraj-bh]

Usually we create folder for a service when there are more than just the service definition file and test file. We can preserve the flat hierarchy here.

[shivaraj-bh]

Any reason why we can't just use Grafana here?

[conscious-puppet]

i copied the doc from here:

• https://grafana.com/docs/grafana/latest/fundamentals/#grafana-open-source

i think, they probably want to distinguish this from Grafana Cloud. I'll change* it to grafana, sure