Urbit Sysops

Contains some helpful scripts for hosting Urbit.

There are two variants included in this repo: one using docker, and one using plain systemd.

The docker approach requires a working docker install, but has the advantage of making it possible to assign predictable HTTP ports per urbit instance; as such it is preferred unless you’re hosting ships that do not require HTTP access, or hosting one ship per VPS.

urbit-exporter

urbit-exporter is a prometheus exporter for urbit. It exports the variables urbit_up and urbit_us (indicating that the login page returned HTTP 200, and that it seems to be for the expected ship, respectively.) Run it by passing the TCP port on which to start its own server (a handy convention may be to use the TCP port corresponding to that urbit’s assigned UDP port, but any scheme is fine), the urbit ship name, and that urbit’s http port. A systemd script is included to automate this.

For example, suppose you have ~sampel-palnet listening on http port 8080/tcp, and you wish to run the exporter on port 42069/tcp. Then you might have:

$ cat /etc/urbit/sampel-palnet.exporter
PORT=42069
SHIP=sampel-palnet
ARGS=--http=8080
$ cat /etc/prometheus/prometheus.yml
[…]
scrape_configs:
  - job_name: urbit
    static_configs:
      - targets: ['localhost:42069']
[…]
rule_files:
  - "/etc/prometheus/rules/*.yml"
$ cat /etc/prometheus/rules/alert-rules.yml
[…]
groups:
  - name: alert-rules
    rules:
    - alert: UrbitDown
      expr: urbit_up == 0
      for: 5m
      labels:
        severity: critical
      annotations:
        summary: 'Urbit down {{ $labels.instance }}'
[…]
$ readlink /etc/systemd/system/multi-user.target.wants/urbit-exporter\@sampel-palnet.service
/etc/systemd/system/[email protected]

urbit-meld

This is intended to be run as a cron job. It stops an urbit instance, runs urbit-worker meld on it, and restarts the instance after either success or failure.

urbit-meld is designed for a systemd-based setup; urbit-meld-docker is designed for a docker-based setup.

urbit-meld includes systemd timer files, so it's sufficient to systemctl enable --now urbit-meld@$ship.timer to get it to run weekly on that ship.

For urbit-meld-docker, you should somehow or another have a list of the ships that you’re running, and just have it do all of them one after another in sequence in a cron.weekly file or something.