Interactive infrastructure map (panel -> cluster -> node -> guest) rendered
with React Flow, backed by a single aggregated /api/network/topology endpoint
instead of fanning out over servers/nodes/SDN/VMs. Adds a netN NIC parser for
QEMU/LXC guest configs, status/threshold filters, canvas tools, and PNG/SVG
export.
The CT template select listed both the raw archive file name and the
headline on one line, which was long and hard to scan. Render the
headline alone and keep the file name in the title tooltip, matching
the convention already used on the images page. Falls back to the file
name when the repository provides no headline.
The dialog body is a grid whose implicit column is auto-sized, i.e.
max-content. Long unbreakable option labels (e.g. CT template file
names) widened that column and pushed the select fields and footer
buttons past the dialog edge.
Use grid-cols-[minmax(0,1fr)] on DialogContent to cap the track at the
dialog width, and apply the same guard to the Alert grid template.
Verified: dialog scrollWidth now equals clientWidth at both 1280px and
375px viewports, with no horizontal page overflow.
Add min-w-0 and overflow-hidden to SelectTrigger and min-w-0 to
SelectValue so long option labels (e.g. CT template names) and grid
columns no longer push select fields past the modal boundary. Fixes
overflow across all dialogs sharing the Select component.
- Surface a topbar indicator with a live count of active tasks
- Propagate exceptions from Proxmox LXC container creation instead of
swallowing them, and validate minimum password length
- Replace the Fedora catalog icon with the official logo artwork
Prevent edit/delete action buttons from overlapping with long template
names by using shrink-0 on icon and button containers with flex-1
min-w-0 on the text area. Make header and filter selects responsive
with flex-wrap and full-width on mobile breakpoints.
Add DNS name validation (RFC 952/1123) for VM, LXC and ISO instance
names on both backend and frontend. Proxmox rejects names with
underscores or special characters but the error was only surfaced as
a raw 400 from the API.
Also require Cloud-Init user and password fields for VM deploys
(highlighted when empty, Next button blocked), and fix numeric input
fields that showed "0" on clear and concatenated values like "04096".
Both toasts printed the endpoint's own message field verbatim, so a Russian UI
answered "Terminated 0 sessions". That string is assembled in Python and can
never be translated where it is built, so the endpoints now also return the raw
count and the phrase is composed on the client.
message is kept: it goes to the audit log and may have other consumers, and
adding a field breaks no one.
Russian needs three forms, and zero does not land in the obvious one — 0 is the
"many" category ("сессий", not "сессии") — hence four keys against English's
two.
Fixes both call sites: the per-user button in the table row and Terminate all
sessions on the Sessions tab, which shared the same untranslated string.
The wizard already refuses to open once a limit has no headroom left, but the
button leading to it still looked available, so the only way to find out was to
click it. Reuse exhaustedMetrics() from the wizard's quota module, so the button
and the wizard cannot disagree about what counts as exhausted, and skip the
check for admins for the same reason the wizard does — their own quota says
nothing about an instance they create for someone else.
The button is rendered as a Link, and disabled does nothing to an <a>: it would
still navigate. So the exhausted case renders a plain button with no Link, and a
tooltip naming the limits that are full. The trigger wraps it in a span because
a disabled button receives no pointer events of its own.
The wizard keeps its own block, which is what still covers navigating straight
to /instances/create.
setting:view used to gate the whole /settings page, which mixed panel-wide
configuration with a user's own account. Revoking it to keep a standard user
out of the panel settings also took away their profile, password, 2FA and SSH
keys, so the permission could not be used for what it names.
The account half now lives at /profile: its endpoints are self-scoped and only
require a session, so the page carries no permission at all and is reached from
the user menu (and Ctrl+K) rather than a sidebar group. /settings keeps the
panel-wide tabs behind setting:view, with saving gated on setting:update,
security on setting:manage and the SMTP/Telegram channels on is_admin, matching
what each endpoint checks.
The shipped "user" role no longer gets setting:view. Migration 42 drops it from
roles holding it without setting:update/manage, since those grants only existed
to reach the profile page; it runs once (marker in panel_settings) so a later
deliberate grant is not reverted on the next start.
Hiding a sidebar entry was the only frontend gate: /users, /logs and
/settings opened for anyone who typed the URL, and their pages rendered
every tab regardless of the caller (the backend then answered the tab's
requests with 403s).
Route elements are now wrapped in RequirePermission, gated on the same
permission the page API requires, so a hidden menu item can no longer be
reached by URL. Denied users land on the first page they can open, or on
an explanatory screen when nothing is reachable, so the redirect cannot
loop. Permission checking moves into lib/permissions.ts and canSeeNavItem
delegates to it, keeping menu and guard in sync.
Settings hides the panel tab without setting:update, the security tab
without setting:manage, the SMTP/Telegram cards from non-admins, and the
update controls behind setting:manage; profile, SSH keys and personal
notification preferences stay available to everyone. Users hides the
roles tab without role:view and the security tab without setting:manage.
Quotas were only enforced by check_quota() on the backend, which answers 429
from the deploy endpoint — so a user walked all five wizard steps and learned
about the limit on the very last click. QuotaHint coloured the offending number
red but blocked nothing.
Split the two situations the UI never distinguished:
* No headroom at all (used >= limit): no configuration would ever fit, so the
wizard does not open. It is replaced by a message naming exactly which
limits are used up, with their numbers.
* Headroom left but the chosen configuration does not fit (used + requested >
limit): the wizard opens, and from the config step onwards it names the
metric and disables Next/Deploy until the resources are reduced.
The overrun check deliberately stays quiet until the config step. Gating every
step meant a default size that did not fit the remaining quota stranded the user
on the server picker, with no way to reach the fields and shrink it.
Neither block applies to admins: their own quota says nothing about an instance
they are creating on behalf of someone else, and blocking on it would lock an
admin with a quota out of the wizard entirely. The figures are still shown.
Also label the quota card as a projection. It renders used + the instance being
created, so a user with 3 of 4 instances saw "4 / 4" and reasonably read it as
already exhausted — which is what prompted this.
The arithmetic lives in features/instances/quota.ts as pure functions rather
than inline in JSX, so the two thresholds can be reasoned about on their own.
Server card used Math.max over all cluster nodes' uptime, so
rebooting one node still showed another node's higher uptime,
making it look like the reboot had no effect.
- Images: show real ISO files already on the node storage (incl. those
present before the node joined the panel) via existing node/isos endpoint
- ISO dialog: rename "detach" to "Eject", show current image inside the
select (drop overflowing side text), add "boot from disk" checkbox
- Backend: on eject, put disk first in boot order; on mount, optional
"boot from this ISO" puts CD-ROM first
- Reboot to apply the new boot order via hybrid restart: graceful ACPI
shutdown with forceStop fallback, then start (fixes guest-ping timeout
on freshly installed OS without QEMU Guest Agent)
- Update iso_next_steps hint to reflect the automated flow
The per-column "Node" filter memoized its option list against TanStack's
Column object, which keeps the same identity across data updates, so the
checklist stayed frozen at whatever nodes were visible when the popover
first mounted instead of reflecting the current (workspace-scoped) rows.
Switching the active workspace also left the server/node/column filters
holding values from the previous workspace, which could filter the newly
scoped VM list down to nothing.
Replace free-text storage inputs with dropdowns populated from the
server/node's actual storages (filtered by vztmpl/rootdir content),
matching the pattern used elsewhere in the app.
Add detect/import flow for nodes already clustered on the Proxmox side
(outside the panel). Matches nodelist entries from cluster/config/join
against known panel servers by IP/hostname and links their panel
records to the cluster_name group, without running any pvecm operation.
Adds a local file upload path alongside the existing URL download flow:
a new /images/upload endpoint streams the file to a temp path and hands
it to a background worker that proxies it into Proxmox storage via the
upload API (requests-toolbelt handles the large multipart stream). Also
adds /images/resolve-filename to recover the real filename (Content-
Disposition) for URL-based downloads with opaque paths.
Raise nginx client_max_body_size and timeouts on both the frontend proxy
and the standalone nginx configs so multi-GB ISOs don't get cut off.
- Invalidate workspace list cache after assigning servers/users so
server/user counts on workspace cards update immediately.
- Invalidate node/VM caches after workspace server/user assignment so
the Nodes page reflects visibility changes right away.
- Soft-delete the cached VM/LXC row and broadcast vm_deleted the moment
a VM/container is destroyed, instead of waiting for the 10s
background sync_vm_cache tick.
- Invalidate the VM list on the frontend when a deploy task completes,
since the backend already persists the new instance to cache
synchronously before marking the task done.
Extract ServerStatusBadge so every server list renders the auth-error
state identically instead of repeating the online/offline ternary, and
expose last_error_kind from the cluster topology and user-assignment
endpoints, which build their payloads by hand and so did not inherit it
from the server schema.
Add a third instance kind to the create wizard: a blank KVM VM with an
empty disk, the installation ISO mounted on ide2 and put first in the
boot order, so the guest boots straight into the installer over noVNC.
Picking a Windows guest presets OVMF, a SATA disk and an e1000 NIC,
which the installer sees without virtio drivers loaded.
Also teach the image downloader content=iso, so ISOs can be fetched
onto a node from the panel: by URL from the images page or from a saved
mirror (mirrors now accept kind=iso), and straight from the ISO step of
the wizard.
The bundled catalog stays qcow2-only on purpose — install ISOs have no
stable "latest" URLs, so hardcoded links would rot on every point release.
An offline server used to report a single "Failed to connect" state, but
"the node does not answer" and "the node answered 401" are fixed in
completely different ways — the first is power or network, the second is a
stale token, which is what happens when a node joins a cluster and its
user.cfg and priv/token.cfg are replaced by the cluster ones.
- add ProxmoxClient.check_connection() returning (ok, error_kind, error);
token auth surfaces 401 as ResourceException rather than
AuthenticationError, since proxmoxer validates the token on first use
- store the reason in proxmox_servers.last_error_kind (migration 40) and
expose it through the server response
- surface an auth_error state in the UI with a hint explaining the usual
cause
Storage listings now also report per-node availability: the cluster config
declares a storage for the whole cluster but says nothing about whether it
actually came up on a given node, so a missing volume group or pool stayed
invisible. Availability comes from a single /cluster/resources call.
Servers can now be connected with a login and password: the panel
authenticates once, creates its own API token on the node (privsep=0) and
uses it from then on. The token can be revoked from the Proxmox UI, cannot
be used for SSH or web login, and appears in the PVE audit log under its
own name. The password stays stored (encrypted) because ProxmoxClient needs
it for SSH fallbacks such as pvecm delnode.
- add proxmox/token_provisioning.py: token creation with free-name
selection and revoke-on-failure so a failed registration does not leave
an orphaned token whose secret was never persisted
- POST /api/servers accepts auto_create_token
- add POST /api/servers/{id}/provision-token to migrate servers that are
still on password auth
- store api_token_value as EncryptedString; it was kept in plaintext while
the password next to it was already encrypted
- drop the unused ProxmoxServerBase schema, a stale copy of the create
schema whose validator rejected the new auth mode
Node detail pages gain a Graphs tab (CPU, IO delay, memory, load average,
network, swap, root filesystem) built on the existing node rrddata
endpoint. The chart card is extracted from the instance graphs tab so both
views share it, and X axis ticks now switch to dates on multi-day ranges.
The overview tab rendered "no data" while the VM list was still loading, so a
server with instances looked empty until the request came back. Gate the empty
state on isLoading and show skeleton rows instead.
Also add the missing nodes.instances_on_server key to both locales — the raw
key was rendered in place of the heading.
Shift dark-theme background, card, popover, sidebar and border/input
tokens from neutral hue 286 (near-zero chroma) to a blue-graphite
hue 258 with higher chroma, so the panel reads as midnight-blue
instead of flat black.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The panel update status lived only in-memory and was cleared solely by the
host watchdog restarting the container. If the rebuild failed mid-way (e.g. the
build OOMs the host when no swap is configured), the container never restarted
and the UI stayed at "restarting - 80%" forever.
- update_service: get_update_status() now self-heals — after UPDATE_TIMEOUT_SECONDS
(default 900s) of "is_updating" it transitions to a terminal "failed" state with
a clear error pointing at the host update service / logs.
- use-settings: poll update status while the server reports is_updating (not only
right after the click in the same tab), so reloading a stuck page keeps refreshing.
- SettingsPage: banner now renders the "failed" state with the error message.
- update_host.sh: build images sequentially (app, then frontend) to lower peak
memory and avoid the OOM that killed the build mid-run.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Frontend read res.count but the auto-import endpoint returns imported_count,
so the toast showed "Импортировано: undefined". Read imported_count and show
a distinct message when no new templates were imported.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Community scripts (build.func) install packages like Java/OpenSearch and
can take minutes, well past the nginx/proxy HTTP timeout. execute_script
now returns immediately with status "running" and finishes the job on a
background thread with its own DB session; the frontend polls
/api/scripts/executions/{id} every 3s while running.
Also set TERM=xterm for script execution over SSH — community scripts
call `clear` and print colored output, which fails with "TERM
environment variable not set" when run without a PTY.
The workspace edit dialog reused useServers(), which always sends the
X-Active-Workspace header, so admins assigning servers to a workspace
only ever saw servers already in the currently active workspace instead
of the full list. apiClient now lets a caller override the header, and
useServers(allWorkspaces) requests '0' (admin "show all") for that dialog.
При редактировании ноды (например, при смене имени) форма отправляла
пустые api_token_name/api_token_value, т.к. секреты не возвращаются
с бэкенда. Бэкенд принимал их как явно заданные и затирал реальные
credentials, из-за чего немедленная проверка подключения падала и
сервер помечался offline. Теперь пустые значения при редактировании
трактуются как «не менять» — и на фронтенде, и на бэкенде.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
При смене токена (вход под пользователем / возврат к админу) сбрасываем
выбранную рабочую область через setActiveWorkspace(null): выбор
предыдущей личности передавался в хедере X-Active-Workspace и мог не
принадлежать новой. Дополняет уже существующий сброс кэша React Query.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Backend:
- POST /api/auth/impersonate/{user_id} — только для админа, выдаёт JWT
с клеймом impersonator; запрещает вход под собой, отключённым,
другим админом и вложенную имперсонацию
- POST /api/auth/stop-impersonate — по клейму impersonator завершает
сессию имперсонации и перевыпускает токен исходного админа
- /api/auth/me отдаёт поле impersonator для баннера в SPA
- аудит-лог: impersonate_start / impersonate_stop
Frontend:
- auth-store: действия impersonate/stopImpersonate со сбросом кэша
React Query при смене токена
- UsersPage: кнопка "войти под пользователем" (только админу,
скрыта для себя/админов/неактивных и во время имперсонации)
- AppLayout: баннер имперсонации с кнопкой "вернуться к администратору"
- локали ru/en
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
При входе под другим пользователем в той же вкладке React Query отдавал
закэшированные данные предыдущего пользователя (список инстансов и др.) до
жёсткой перезагрузки, из-за staleTime и отсутствия сброса кэша при смене сессии.
- queryClient вынесен из App.tsx в lib/query-client.ts (доступен вне React)
- auth-store вызывает queryClient.clear() при успешном login и при logout
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Свёрнутые долгие операции (установка приложения, сборка золотого шаблона)
теперь живут в глобальном zustand-сторе (persist в localStorage) и
рендерятся треем MinimizedOpsTray в AppLayout, а не самим диалогом:
- плашка переживает навигацию между страницами и перезагрузку вкладки;
прогресс опрашивает сам трей (REST, 3с) независимо от диалога
- клик по плашке возвращает на страницу операции и разворачивает диалог:
установка — resume-режим InstallWizard (resumeInstall в navigation
state), шаблон — openGolden с предвыбранным сервером
- «Свернуть» добавляет операцию в трей и закрывает диалог; крестик на
плашке только убирает её (операция продолжается на сервере)
- плашки несуществующих операций (404 после перезагрузки) трей убирает сам
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Доставка файлов приложения из каталога (data_files) до compose up,
APP_PASSWORD/default_credentials для Umbrel, сворачивание диалогов
долгих операций (MinimizableDialog), фильтрация pull-шума в ошибках.
Исправления багов:
- teardown: отсутствующий в Proxmox CT считается удалённым — orphaned-приложение
теперь можно удалить из БД
- install: сдвиг VMID мимо занятых записей (nextid не резервирует номер,
retry мог подхватить чужой контейнер)
- retry: параметры установки сохраняются в installed_apps.install_params
(cores/memory/disk/storage/bridge/ip_config/port_explicit, миграция JSONB)
вместо жёстких дефолтов
- reconcile: недоступный сервер больше не помечает все приложения orphaned
- update: реально опубликованный порт фиксируется в БД (url/health)
- .env: значения с '#' кавычатся (dotenv-парсер обрезал их как inline-комментарий)
- порт 80, выбранный явно в мастере, публикуется как есть
- useCatalog: available_only включён в queryKey (коллизия кэша)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
App Store install/update pipeline reliability fixes:
- explicit nameserver (APPSTORE_NAMESERVER) on App Store LXCs and the golden
template build CT, instead of inheriting a node resolver that may be
unreachable from the container network (e.g. Tailscale-only DNS)
- golden template storage/resolution moved from one global setting to
per-server_id (migration 37: app_operations.server_id) — building it on one
Proxmox server no longer overwrites the template configured for another
- golden template automated build fixed: vzdump export step no longer fails
on "unable to parse directory volume name 'vztmpl/probe'"
- bind-mount host directories are pre-created with permissive ownership
before `docker compose up`, fixing Permission denied for non-root
containers (e.g. MariaDB) writing to Docker-auto-created root-owned dirs
- delete/teardown now waits for Proxmox to actually confirm the destroy task
finished (was silently proceeding after a short timeout, letting a new
install reuse the VMID while the node was still wiping the old disk)
- pct create retries a few times on transient "got lock" config errors
- failed pipeline steps now report the tail of command output (4000 chars)
instead of the head, so the real error isn't buried under pull/progress noise
- install error box in the UI wraps long text and scrolls instead of
stretching the dialog
Also carries forward in-progress script catalog / instances / RBAC changes
that were already uncommitted in the working tree before this session.
Bump to v1.9.1, update WIKI.md/docs/CHANGELOG accordingly.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Командная палитра (Ctrl+K) с общим реестром nav-items и его store
- Новые общие UI-примитивы: alert, breadcrumb, checkbox, progress, switch, textarea
- Меню инстанса: добавлены «Параметры», «Мигрировать», «Изменить ресурсы»,
«Создать резервную копию» (новый BackupDialog) и «Удалить»; исправлено
закрытие меню при живом обновлении списка (стабильные columns + getRowId)
- Вкладки инстанса: «Вычислительные ресурсы», «Опции CPU», «Управление
дисками» и «Изменение размера диска» перенесены с «Настройки» на
«Оборудование» для VM (для LXC, у которых нет вкладки «Оборудование»,
ресурсы и ресайз диска остаются в «Настройках»); общие карточки вынесены
в ResourceCards.tsx
- Локализация: новые ключи ru/en для добавленных пунктов меню и бэкапа
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>