counterweight
6c1bcbed95
552-line playbook becomes 18 lines plus a 411-line role
(install/service/healthcheck phases, four templates, two handlers).
phoenixd_vars.yml is deleted; its content is the role's defaults.
Task-list diff vs the old playbook shows ONLY the eight Uptime Kuma tasks
removed - everything else identical and in the same order.
phoenixd holds a Lightning node, so the run was checked against a pre-flight:
before: channel 6c25fa83..., balanceSat 1550723, capacitySat 3114830,
blockHeight 966692, active since 2026-09-02
after: identical, and still active since 2026-09-02 - it did NOT restart
`Create phoenixd systemd service` came back unchanged, which is what proves the
template reproduces the live unit byte-for-byte. changed=3 was the health check
script, its unit (Environment rename), and the timer restart. Second run:
changed=0.
Two things the conversion fixed, both symptoms of the deprecation banner having
been applied to contiguous blocks rather than to individual tasks:
- The health check logged "ERROR: UPTIME_KUMA_PUSH_URL not set" on every fire -
about 1,400 times a day - because its Environment= was emptied at
decommissioning. The exit code was still correct so nothing was broken, but it
is exactly the kind of noise that trains you to ignore a log. An unset push
URL is now normal and silent.
- `Enable and start phoenixd health check timer` was guarded by
uptime_kuma_enabled and so had not run since the decommissioning, while the
timer itself was still live on the host from before. Ansible had quietly
stopped managing something that was still running. Ungated.
Noted, not changed: seed.dat is mode 0644 on the host. That is phoenixd's own
doing, but it is a Lightning seed and worth tightening.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
56 lines
2.3 KiB
Markdown
56 lines
2.3 KiB
Markdown
# `phoenixd`
|
|
|
|
Deploys and runs [phoenixd](https://phoenix.acinq.co/server), an ACINQ Lightning
|
|
node, on the edge host. LNBits uses it as a wallet backend. The HTTP API stays on
|
|
loopback — phoenixd is never published through Caddy.
|
|
|
|
Converted from `deploy_phoenixd_playbook.yml` (552 lines) under Plan 6. The
|
|
playbook is now 18 lines.
|
|
|
|
## Phases
|
|
|
|
| | |
|
|
|---|---|
|
|
| `install.yml` | packages, system user, directories, versioned download and install |
|
|
| `service.yml` | systemd unit, start, then first-boot checks (config written, seed created) |
|
|
| `healthcheck.yml` | check script, unit, timer |
|
|
|
|
## The seed
|
|
|
|
`{{ phoenixd_data_dir }}/seed.dat` **is** the funds. phoenixd is deliberately
|
|
excluded from the automated backups (Plan 5, Model C): the seed is twelve fixed
|
|
words that never change, so an automated job would only manufacture more copies
|
|
of a static secret on more machines. Write them down offline, once.
|
|
|
|
Note the live file is mode `0644`. That is phoenixd's own doing, not this role's,
|
|
and it is worth tightening.
|
|
|
|
## Monitoring: one variable, no product knowledge
|
|
|
|
The check asks the node itself — the service must be active **and**
|
|
`phoenix-cli getinfo` must return a `nodeId` — and records the answer in its exit
|
|
code, which systemd keeps:
|
|
|
|
```bash
|
|
systemctl is-failed phoenixd-healthcheck.service
|
|
```
|
|
|
|
That is a complete answer with no monitoring system involved. To report
|
|
elsewhere, set `healthcheck_push_url` to anything accepting an HTTP ping. Gone
|
|
from this role: the embedded Python that created monitors over the Uptime Kuma
|
|
API, the `/tmp` credentials file, the push-URL file written and parsed back, and
|
|
the systemd `Environment=` rewrite.
|
|
|
|
### Two things the conversion fixed
|
|
|
|
**The check used to log an error once a minute.** Its `Environment=` push URL had
|
|
been empty since the decommissioning, and the script printed
|
|
`ERROR: UPTIME_KUMA_PUSH_URL not set` on every fire — roughly 1,400 times a day.
|
|
The exit code was still correct, so nothing was broken; it was pure noise, and
|
|
noise that trains you to ignore the log. An unset push URL is now normal and
|
|
silent.
|
|
|
|
**`Enable and start phoenixd health check timer` was guarded by
|
|
`uptime_kuma_enabled`** and so had not run since the decommissioning — while the
|
|
timer itself was still live on the host from before. Ansible had quietly stopped
|
|
managing something that was still running. Ungated.
|