Files
EvoFirewall/docs/architecture.md
T
Denozordec d552f4f326
Build and Push EvoFirewall Docker Image / build-and-push (push) Successful in 1m45s
Build and Push EvoFirewall Docker Image / create-release (push) Skipped
feat(api): implement self-update mechanism for Linux agents
- Added a `maybe_self_update` function in `evofw-firewall.sh` to allow agents to pull the latest version of the sync script from the server, enhancing the agent's ability to stay updated.
- Updated the `/v1/agent/sync-script` endpoint to return ETag and script SHA256 headers, enabling efficient caching and conditional requests.
- Modified the agent policy response to include `script_sha256`, providing visibility into the current version of the sync script.
- Enhanced tests to verify the self-update functionality and ensure correct behavior of the sync script endpoint.

These changes improve the maintainability and reliability of Linux agents by enabling automatic updates of critical scripts.
2026-08-15 16:56:00 +07:00

39 lines
2.3 KiB
Markdown

# Архитектура EvoFirewall
Централизованный control plane для firewall-агентов (Linux nft/ipset, MikroTik address-list).
## Компоненты
| Компонент | Путь | Роль |
|-----------|------|------|
| Web SPA | `apps/web` | ReUI Frame, TanStack Router/Query |
| API | `apps/api` | Fastify 5, JWT + agent tokens |
| DB | `packages/db` | Drizzle + SQLite WAL |
| Shared | `packages/shared` | Zod-контракты, RBAC helpers |
| UI | `packages/ui` | shadcn primitives `@evofw/ui` |
| Agents | `apps/api/src/agent-scripts` | install.sh, sync, MikroTik RSC |
## Потоки
1. **Enroll**`POST /v1/agent/enroll` + `X-EvoFW-Seed` → pending agent
2. **Approve** — UI/API → status approved
3. **Policy**`GET /v1/agent/policy` → deny/allow CIDRs + `default_action` + optional `port_rules` + hash (`apply_version: 3`) + `script_sha256` (Linux; не в `policy.hash`)
4. **Linux self-update** — timer: `GET /v1/agent/sync-script` (`ETag` / `If-None-Match`) → при новой версии заменить `/usr/local/sbin/evofw-firewall.sh` и `exec` до policy
5. **Apply** — agent пишет kernel rules (L3 + L4 port ACL на nft), `POST /v1/agent/apply-report` + stats + optional `host_firewall` snapshot
6. **Lists refresh** — cron каждые 5 мин (json_url / domains / evobgp_community)
## Политика
- Именованные **наборы правил** (`policy_sets`); агенту назначается **M:N** через `agent_policy_sets`
- Правило в наборе: `action: deny | allow` + ровно один источник — IP-список (`list_id`), CIDR или DNS-имя (`hostname` → A/AAAA, кэш в `policy_rule_resolved`)
- Evaluate: правила всех назначенных enabled-наборов (sort + priority) + `ip_overrides`
- Цепочка ядра **всегда**: deny → allow → `default_action` (`accept` | `drop` на агенте)
- На Linux nft: после allow — **Port ACL** (`close` drop, затем `open` accept) из `agent_port_rules`
- Exact overlap: `allow \ deny` (`conflicts_dropped`); deny wins
- Overrides, смена наборов, `default_action`, Port ACL и refresh DNS/lists бампят `policy_generation`
## Auth
- Portal SSO app id **`fw`**, permissions `fw:*`
- Agent bearer token (sha256 hash в БД)