Dashboard

The dashboard updates itself: a scaled service or a downed node appears without a refresh. The connection indicator in the nav bar shows when the last update arrived, and reconnects on its own if the connection drops. It also carries a resync button that makes Cetacean re-read the whole cluster from Docker.

The indicator reads Live, Reconnecting, or Stale. Stale means your browser is connected to Cetacean but Cetacean has lost Docker, so everything on screen is the state it last read, not the cluster as it is now. The indicator switches to a warning color and shows how long ago that was; the dashboard catches up on its own once Docker answers. The same condition is available to alerting—see Monitoring.

Navigation

The nav bar links every resource type: nodes, stacks, services, tasks, configs, secrets, networks, volumes, plugins, plus the swarm info, topology, and metrics pages. On narrow screens they collapse behind a menu button.

To the right of the search box sit the shortcut help button, a recommendations indicator badged with the current finding count, the theme toggle (light, dark, system), and, in every authentication mode except none, the identity badge linking to your profile. The footer carries the running version and commit, the licenses page, and a link to the API playground at /api.

Keyboard shortcuts

Press ? for the full list. Shortcuts are ignored while focus is in a text input.

⌘ K / Ctrl K

Toggle the command palette

/

Open the command palette

?

Toggle the shortcut list

Esc

Close the overlay, or go back

g h

Cluster overview

g n

Nodes

g k

Stacks

g s

Services

g a

Tasks

g c

Configs

g x

Secrets

g w

Networks

g v

Volumes

g p

Plugins

g i

Swarm info

g t

Topology

g m

Metrics console

g r

Recommendations

j / ↓

Next row in a table

k / ↑

Previous row

Enter

Open the selected row

The g shortcuts are chords: press g, release, then the second key within one second. Hovering a nav link shows its chord.

Command palette

⌘ K (Ctrl K on Linux and Windows) opens the palette. Type to search names, images, and labels across every resource type. Results are grouped by type in a fixed order, with a state indicator per row, and refresh every two seconds while the palette is open so a converging service updates in place. Move with the arrow keys and open with Enter.

Typing an action name instead runs that action as a guided sequence, one prompt per argument, with a breadcrumb of what you have chosen so far:

Scale Service

Pick a service, then a replica count

Update Image

Pick a service, then an image reference

Rollback Service

Pick a service

Restart Service

Pick a service

Drain Node

Pick a node

Pause Node

Pick a node

Activate Node

Pick a node

Promote Node

Pick a node

Demote Node

Pick a node

Force Remove Task

Pick a task

Remove …

Pick a service, node, stack, config, secret, network, or volume

Actions that remove or disrupt something ask for confirmation before they run. An action your operations level or your grants don’t allow is refused with a permission message, not a server error.

List pages

Every resource type has a list page with a search box, sortable columns, and a table or grid toggle. Search and sort are held in the URL (?q=, ?sort=, ?dir=), so a filtered list is a shareable link. Your choice of table or grid is remembered per resource type; narrow screens always use the grid.

Lists load more as you scroll, and stay current as resources come and go.

Where your grants allow creating one, the config and secret lists carry a create form: a name, plus the content pasted in or read from a file. A secret’s value can’t be read back afterward.

The API accepts expression filters through ?filter= that the dashboard’s search box doesn’t build. See the API guide for the fields available per resource type.

Detail pages

A detail page shows the resource, its cross-references (the services using a config, the tasks of a service, the stack a resource belongs to), and its recent change history. Cross-references are links, so you can walk from a secret to the services mounting it to the nodes their tasks run on.

A stack’s page opens with Topology: the networks its services attach to and the configs, secrets, and volumes they mount, drawn as a graph above the tables listing the same resources. Every node links to the resource it stands for, and a service shows its running tasks against the number wanted. A network the stack doesn’t own is drawn dashed; a resource nothing references is dimmed. Hovering or focusing a node dims everything more than one edge away and names it in the URL.

Where the operations level and your grants allow it, detail pages carry actions: scale, update image, rollback and restart on a service, plus inline editors for environment variables, resource reservations and limits, placement, ports, update and rollback policy, and the log driver; availability and labels on a node; force removal on a task. Actions hidden by permissions aren’t rendered.

Configs, secrets, networks, volumes, services, stacks, nodes, and tasks can also be removed from their detail page, behind a confirmation. A config, secret, network, or volume that a service still uses can’t be removed, and its button says why.

Service and stack pages end with a collapsed Compose file section. Expanding it renders the resource as a compose document, with buttons to copy it or download it as a .yaml file. It redeploys to the state the cluster is running now, so it isn’t the file the stack was first deployed from, and it references secrets and configs rather than exporting their contents.

Rotating a secret or config

Swarm can’t change a secret or a config once it exists, so rotating one means replacing it:

1. On the Secrets list, create the replacement under a new name, such as db_password_v2. This needs operations level 2.
2. Open the old secret. Its page lists every service that mounts it. On each service, edit Secrets: add the replacement with the same Target the old one had, so the container reads the same file, and remove the old one. Saving rolls the service out once. This also needs level 2.
3. Once the old secret’s page lists no services, remove it. This needs level 3, and Docker refuses while any service still mounts it.

Configs follow the same steps from the Configs list and each service’s Configs editor. A service deployed from a Compose file gets the old reference back on the next docker stack deploy, so change the file as well.

Over the API, the same sequence is POST /secrets, then PATCH /services/{id}/secrets for each service named in GET /secrets/{id}, then DELETE /secrets/{id}; see the API guide.

Charts

Charts appear on the cluster overview and on node, service and task detail pages, and require monitoring. Node and service list pages carry sparklines and gauges from the same data.

The panel header holds a 1H / 6H / 24H / 7D selector, a custom date-time range picker, a refresh button, a pause control for the live stream, and a line/area toggle. The selected range is stored as ?range=, and a custom one as ?from= and ?to=.

• Click a series name or line to isolate it. Everything else dims; click again to restore.
• Drag horizontally to zoom into a time window. The URL updates, so the window is shareable.
• Hover one chart to get a crosshair and matching values on every other chart in the same panel.
• Double-click a stack in the per-stack charts on the cluster overview or a node page to drill into its services.

The preset ranges update live as new points arrive. A custom range is a snapshot; use the refresh button to bring it up to date.

Metrics console

The metrics page runs PromQL directly against the configured Prometheus, with completion for metric names. A query returns a result table and a range chart over the same 1H / 6H / 24H / 7D selector. The query and range are held in the URL as ?q= and ?range=.

Log viewer

Service and task detail pages carry a log viewer. The live toggle in the toolbar starts and stops the stream; while it runs the view follows the bottom of the output, and scrolling up pauses that until you scroll back down.

• Time range: presets from the last 5 minutes upwards, or a custom since/until pair
• Filters: by level (parsed from the line, including JSON and log/slog numeric levels) and by stream
• Search: substring or regular expression, with match navigation and highlighting
• Rendering: JSON payloads pretty-printed, levels color-barred
• Download: saves the lines currently loaded as a .log file

Topology

The topology page has two views.

• Logical: one card per service, grouped into its stack, with an edge between any two services sharing an overlay network, listing the shared networks. Hovering a card dims everything it isn’t connected to, and the legend maps colors to stacks.
• Physical: one card per cluster node, listing the tasks placed on it.

Both views pan, zoom and let you drag cards, and clicking a service opens its detail page.

Swarm page

The swarm info page shows the cluster’s join tokens and its Raft, orchestration, dispatcher, CA, and encryption settings, each editable in place where the operations level and your grants allow it. Rotating a join token asks for confirmation.

Below the settings, a Plugins section lists the installed Docker plugins and links each to its own page, where you can enable, disable, configure, upgrade, or remove it. An Install Plugin button appears when your grants allow installing. It takes a remote reference and shows the privileges the plugin requests before you confirm.

Profile

In every authentication mode except none, the identity badge opens a profile page. It shows how you’re signed in, with your subject, email, and groups where the provider supplies them, and the certificate issuer and expiry in cert mode. Under oidc it carries the Sign out button.

Installing the dashboard

Cetacean serves a web app manifest, so a browser that supports it offers to install the dashboard as an app with its own window and icon. It’s the same dashboard—it still needs a connection to Cetacean, and nothing is cached for offline use.

History

The history page at /history, linked from the recent activity on the cluster overview, lists the latest 200 changes Cetacean has seen. Filter it by resource type, or by name with the search box. The download button saves the whole log Cetacean still holds as CSV.

Atom feeds

Every resource list and detail page shows a feed icon in the page header. Click it to open that page’s Atom feed, or copy the URL into a feed reader. The history, search and recommendations pages have feeds too. See the API guide for the supported endpoints and pagination.

Licenses

The licenses page, linked from the footer, lists every open-source dependency bundled into Cetacean, both Go modules and frontend packages. Search by name, or filter by ecosystem and license. Clicking a license badge opens the full text for that dependency, with its NOTICE file if it ships one. The header links to the complete attribution document.