Repository navigation
v2.0.0
v2.0.0 — мониторинг в прямом эфире
English · Русский
Дата: 2026-10-18. Предыдущий выпуск: v1.6.0.
Русский
Что нового
- Мониторинг. Новый экран «Мониторинг» показывает вживую, как живёт сервер и сколько трафика идёт по каждому протоколу: CPU, память, диск, сеть, контейнеры, а у MTProxy, NaiveProxy, Mieru, AmneziaWG и TrustTunnel — скорость ↑/↓, число сессий и самых активных клиентов. Он виден всем ролям и ничего не меняет. Подробно — MONITORING.ru.md.
- Три вкладки. «Сейчас» — прямой эфир с графиком за 5 минут; «Трафик» — история по протоколам от 15 минут до 90 дней; «Узлы» — сравнение всех узлов, оповещения и разбор выбранного узла. История хранится до 400 суток (минутные точки 3 суток, 15-минутные 35 суток, часовые 400 суток). Оповещения: CPU, память, диск, подкачка, молчащий узел, всплеск сети.
- Анализ по узлам. Центральная панель показывает и связанные узлы: выбор узла и «Все узлы». Узлы опрашиваются, только пока кто-то смотрит экран. Узел старее 2.0 помечен как
unsupported— сначала обновите узлы, потом центр. - Как это измеряется. Агент обновления снимает показания раз в секунду. Байты по протоколам считают счётчики ядра в отдельной таблице nftables
inet proxy_control_metrics: она только считает и никогда не блокирует и не меняет трафик. Направление — с точки зрения клиента (↑ от клиентов, ↓ к клиентам), это байты сетевого уровня, вместе с заголовками пакетов. Еслиnftнедоступен, панель берёт счётчики менеджеров: NaiveProxy тогда не показывает скорость вживую. Через менеджер MTProxy виден только суммарный счётчик, а у NaiveProxy и Mieru вживую нет скоростей по клиентам. - MCP. Новые инструменты
get_metrics_live,get_metrics_history,get_metrics_summary; вoverview— компактный разделmetrics; навыки утреннего обзора и диагностики доступа ими пользуются. - Обновление менеджеров. Агент обновления теперь сам пересобирает менеджеры (
mieru-manager,naive-manager,xray-router,awg-manager,trusttunnel-manager,mcp), которые изменило обновление панели, и снимает устаревшую пометку «ещё не пересобраны» на экране «Версии». На сервере с AmneziaWG или TrustTunnel пересборка их менеджеров ненадолго обрывает туннели и сессии; приложения переподключаются сами. У владельца есть кнопка «Пересобрать менеджеры». - Счётчики в меню заполняются при запуске панели, а не только после захода на экран.
Что учесть
- Обновление — обычный
--update. Ломающих изменений нет: API, файлы конфигурации и TOML установщика прежние, настраивать ничего не нужно. Установщик в этом выпуске изменён одной правкой: свежая установка с Mieru теперь задаёт MTU 1400 в начальном конфиге (mita 3.38 сообщает незаданный MTU как 0, и менеджер Mieru его отвергал). Установленные узлы это не затрагивает. - После обновления агент создаёт таблицу nftables
inet proxy_control_metricsи сам находит порты протоколов. Нестандартные порты задаются файлом/etc/proxy-control/metrics-ports.json; убрать таблицу можно командойsudo nft delete table inet proxy_control_metrics. - Миграция базы 25 добавляет таблицу истории. Образ 1.x не работает с базой этой схемы, поэтому откат делайте вместе с базой из резервной копии.
- Скорости считаются по счётчикам сетевого уровня и поэтому немного больше «чистой» нагрузки. Сэмплер занимает порядка 0,3 % одного ядра.
PANEL_METRICS_ENABLED=0выключает мониторинг.
Установка на новый сервер
Требования: Ubuntu 24.04 LTS x86-64, собственные DNS-имена и обычный пользователь с sudo. Не запускайте мастер непосредственно от root. Подробно — установка и справочник.
В пустом рабочем каталоге скачайте файлы именно v2.0.0:
curl -fsSLO https://github.com/dubr1k/proxy-control/releases/download/v2.0.0/install-release.sh
curl -fsSLO https://github.com/dubr1k/proxy-control/releases/download/v2.0.0/install-release.sh.sha256
sha256sum --check install-release.sh.sha256
bash install-release.shПродолжайте только после успешной проверки SHA-256. Мастер спросит домены и состав установки.
Обновление существующей установки
Сначала сделайте резервную копию. Скачайте и проверьте два файла командами выше, затем вместо запуска мастера выполните:
bash install-release.sh --updateВ парке связанных панелей штатный порядок — узлы, затем центр. Перезапуск агента обновления в конце обновления подхватывает измерения; первые данные появляются через несколько секунд. После обновления проверьте состояние служб и откройте «Мониторинг». Подробнее — обновление до 2.0. Не удаляйте контейнерные тома, базу или секреты ради обновления.
English
What is new
- Monitoring. A new «Мониторинг» (Monitoring) screen shows live how the server is doing and how much traffic each protocol carries: CPU, memory, disk, network, containers, and for MTProxy, NaiveProxy, Mieru, AmneziaWG and TrustTunnel the ↑/↓ rate, the number of sessions and the busiest clients. Every role can see it, and it changes nothing. Details: MONITORING.en.md.
- Three tabs. «Сейчас» (Now) is the live feed with a 5-minute chart; «Трафик» (Traffic) is history by protocol from 15 minutes to 90 days; «Узлы» (Nodes) compares all nodes, shows alerts and drills into a chosen node. History is kept up to 400 days (1-minute points for 3 days, 15-minute points for 35 days, hourly points for 400 days). Alerts: CPU, memory, disk, swap, a silent node, a network spike.
- Analysis by node. The central panel shows linked nodes too, with a node selector and «All nodes». Nodes are polled only while someone watches the screen. A node older than 2.0 is marked
unsupported— update the nodes first, then the central panel. - How it is measured. The update agent samples the host once a second. Per-protocol bytes come from kernel counters in a nftables table of its own,
inet proxy_control_metrics, which only counts and never blocks or changes traffic. Direction is the client's point of view (↑ from clients, ↓ to clients), and the bytes are network-layer bytes, packet headers included. Ifnftis unavailable the panel uses the managers' counters, and NaiveProxy then has no live rate. Through MTProxy's manager only a combined counter is visible, and NaiveProxy and Mieru have no live per-client rates. - MCP. New tools
get_metrics_live,get_metrics_historyandget_metrics_summary;overviewgains a compactmetricssection; the morning-overview and access-diagnosis skills use them. - Manager updates. The update agent now rebuilds the managers (
mieru-manager,naive-manager,xray-router,awg-manager,trusttunnel-manager,mcp) that a panel update changed, and clears the stale «not rebuilt yet» note on the «Versions» screen. On a server with AmneziaWG or TrustTunnel, rebuilding their managers briefly drops tunnels and sessions; the apps reconnect by themselves. The owner also has a «Пересобрать менеджеры» button. - Sidebar counters fill in when the panel starts, not only after a visit to the screen.
Good to know
- The update is the normal
--update. There are no breaking changes: the API, the configuration files and the installer TOML are as before, and there is nothing to configure. The installer has one fix in this release: a fresh installation with Mieru now sets an MTU of 1400 in the bootstrap config (mita 3.38 reports an unset MTU as 0, and the Mieru manager refused it). Installed hosts are not affected. - After the update the agent creates the nftables table
inet proxy_control_metricsand finds the protocol ports itself. Non-standard ports are set in/etc/proxy-control/metrics-ports.json; the table can be removed withsudo nft delete table inet proxy_control_metrics. - Database migration 25 adds the history table. A 1.x image does not work with a database of this schema, so roll back together with the database from the backup.
- Rates come from network-layer counters, so they are slightly higher than the pure payload. The sampler takes about 0.3 % of one core.
PANEL_METRICS_ENABLED=0switches monitoring off.
Fresh installation
Use Ubuntu 24.04 LTS x86-64 and a regular user with sudo, not a root login. Prepare your DNS names and follow the installation guide and installer reference.
In an empty working directory:
curl -fsSLO https://github.com/dubr1k/proxy-control/releases/download/v2.0.0/install-release.sh
curl -fsSLO https://github.com/dubr1k/proxy-control/releases/download/v2.0.0/install-release.sh.sha256
sha256sum --check install-release.sh.sha256
bash install-release.shContinue only after checksum verification succeeds. The wizard asks for domains and components.
Upgrading
Make a backup, download and verify the same files, then run:
bash install-release.sh --updateThe normal fleet update order is nodes first, central panel last. The update agent's restart at the end of the update picks up the measurements; the first data appears within a few seconds. Check service health afterwards and open «Мониторинг». More: upgrading to 2.0. Do not delete Docker volumes, the database or secrets as part of an update.