Files
Germaine Brown 98073dc636
Publish Docs Site / build (push) Failing after 5s
feat: initial itpp-docs setup with MkDocs Material
- mkdocs.yml with dark slate theme, nav for 12 ITPP projects
- build-docs.sh aggregates docs from all project repos
- .gitea/workflows/docs-publish.yml for nightly rebuild+deploy
- README and CHANGELOG for the itpp-docs repo itself
- docs-source/ populated from all 12 repos
- site/ ready for deployment to docs.itpropartner.com
2026-08-09 23:49:45 -04:00

113 lines
4.4 KiB
Markdown

# Ops Portal — Architecture
## Topology
```
INTERNET
|
[Caddy :443]
|
Core (152.53.192.33)
|
+---------------+---------------+
| | |
/api/* :8090 /data/* :files /static/*
| | |
[FastAPI app] ops-status.json [HTML/CSS/JS]
server.py /var/www/ops/ /opt/ops-portal/
| /data/ static/
|
+-------+-------+-------+-------+
| | | | |
S3 API UISP Wazuh Bitdef systemd
(Wasabi) (FFW) (app1) (Cloud) (Core)
```
## Data Flow
```
[Collector] [Dashboard]
| |
|-- python3 ops-data- |
| collector.py |
| |
v |
S3 buckets ----+ |
UISP API ------+---> ops-status |
Wazuh API -----+ .json ------> GET /api/status
Bitdefender ---+ |
systemd -------+ |
cron jobs -----+ |
v
[Browser renders
health grid,
widgets, alerts]
```
## Components
### 1. Collector (`/root/.hermes/scripts/ops-data-collector.py`)
- Runs every 5 min via cron
- Gathers: S3 backup status (6 buckets), UISP devices (90), Wazuh agents/alerts, Bitdefender endpoints, systemd services, cron jobs, server health, disk/memory/CPU
- Timeout: 90s (was 20s — too short for 94K-file S3 bucket)
- Output: `/var/www/ops/data/ops-status.json`
### 2. Backend (`/opt/ops-portal/server.py`)
- FastAPI on port 8090
- 7 API endpoints (health, status, servers, servers/health, audit-log, ft360/status)
- JWT auth from `/root/.hermes/.env` (ADMIN_USERNAME, ADMIN_PASSWORD, JWT_SECRET)
- Critical service restart protection (hermes, caddy, ops-portal blocked)
- Systemd: `ops-portal.service`
### 3. Frontend (`/opt/ops-portal/static/`)
- 11 HTML pages with shared ops.css, app.js, utils.js
- Auth: login overlay → localStorage JWT → all API calls Bearer
- Auto-refresh: 60s interval + tab visibility API
- Mobile: hamburger toggle with .nav-links.open CSS
- Cache-busting: all assets versioned with timestamps
### 4. Proxy (Caddy on Core)
- `/` and `/*.html` → static file server from `/opt/ops-portal/static/`
- `/api/*` → reverse_proxy to 127.0.0.1:8090
- `/data/*` → file server from `/var/www/ops/data/`
- Domain: ops.itpropartner.com
## Cross-Service Dependencies
| Dependency | Server | Purpose | Fallback |
|---|---|---|---|
| Wasabi S3 | External | Backup bucket status | Shows "Issues" |
| UISP API | unms.forefrontwireless.com | Device/site count | Shows 0 devices |
| Wazuh | app1 (152.53.36.131) | Agent count, alerts | Shows "Offline" |
| Bitdefender | External API | Endpoint monitoring | Shows "Offline" |
| Traccar | app2 (152.53.39.202) | FleetTracker data | Dedicated endpoint |
| Core systemd | Local | Service health, disk, memory | N/A (local) |
## Auth Flow
```
Browser Server
| |
|-- POST /api/auth/login ->|
| {username, password} |
| |-- Validate against ADMIN_USERNAME/ADMIN_PASSWORD
| |-- Generate JWT with JWT_SECRET
|<- {access_token} --------|
| |
|-- GET /api/status ------->|
| Authorization: Bearer |
| |-- Verify JWT
| |-- Read ops-status.json
|<- {full dashboard} ------|
```
## Key Design Decisions
1. **Collector pattern over direct API calls:** Dashboard fetches one JSON blob rather than 6 separate APIs. Single point of failure but fast rendering and offline-capable (shows last-cached data).
2. **Python/FastAPI over Node:** Already have Python toolchain on Core. FastAPI is lightweight, async-native, and the ops portal is read-heavy with minimal write paths.
3. **Static HTML + vanilla JS over React/Vue:** 11-page dashboard with no SPA routing. Auth via localStorage JWT. Zero build step, zero dependencies beyond ops.css.
4. **JWT over session cookies:** Cross-page auth without server-side session state. Token survives page navigations and ops-portal restarts (persistent JWT_SECRET in .env).