Dashboard Guide

Wiki home

Documentation index · Source repository

Open the Controller address from sh scripts/swarm.sh access and log in with the configured account. This guide covers live panels and controls. Use experiment execution and recovery to run work, saved results to preserve it, and topology to interpret nodes and edges.

Built-in Dashboard visualization

Use the tabs beneath Metrics to switch between Network Overview, Exp Progress, Agent Status, Recent Network Events, and Saved Results. One workspace is displayed at a time; the metrics row remains available above every tab. Keyboard users can use Skip to dashboard to move past the header.

The header's Tools menu contains Prometheus, Grafana, and Log out. Secondary actions are grouped in menus such as More. Open a menu with Enter/Space; Up/Down and Home/End move between available actions. Escape closes it and returns focus to its button. Clicking outside also closes it.

Metrics places all 11 cluster, delivery, bandwidth and observation cards in one horizontal row. Scroll sideways or use the Previous metrics / Next metrics arrow buttons. Hover over a card to expand that card horizontally and reveal its detailed values and explanation; only one card expands at a time. Long details scroll vertically within the expanded card. On narrow mobile screens, expanded details appear below the metric value to keep the text readable.

• Click or tap a card to open its details, and select it again to close them. On touch screens, swipe horizontally to move through the row.
• Keyboard focus opens a card. Use Left/Right to move between cards and Home/End for the first/last card. Enter/Space toggles details; Escape or a click/tap outside the row closes them.
• How delivery is measured remains directly below the row. Its explanation and expanded state persist through live updates.

Online Agents and Ready Peers describe the whole cluster; message and bandwidth cards follow the run shown by Run metrics. Observation quality shows receipt, continuity and start unknown counts separately as Receipt / Continuity / Start; these overlapping categories are not summed. Expand it for the full labels and outcome counts.

When a run finishes, fails or is canceled, its carousel measurements reset to the initial N/A/zero values. All cards remain available; the next running run supplies new values. Cluster counts stay live, and final measurements remain in Saved results.

Available slots totals the free Peer capacity reported by online Agents; offline Agents contribute no available slots. Capacity is an admission count, not a CPU or memory reservation. See Peer placement and capacity before increasing it.

Saved results → Images submits server background analysis and provides overview, message and repeated-run charts as PNG/CSV/ZIP. It uses saved records independently of Prometheus retention. See result images for operation and experiment metrics for definitions.

Workspace tabs

Tab	Contents
Network Overview	Live topology, layers, camera controls, and Peer details
Exp Progress	Runs, phase/job progress, finish estimates, and stop controls
Agent Status	Agent health, capacity settings, and metrics links
Recent Network Events	Recent Peer activity and delivery events
Saved Results	Search, recovery, downloads, analysis, and deletion

Click or tap a tab to open its full-width workspace. With a tab focused, use Left/Right arrows to switch tabs or Home/End to select the first/last tab; Tab moves into the selected panel. On small screens, the tab bar scrolls horizontally. Panel Collapse/Expand controls are no longer used.

The URL records the selected tab (#network, #experiments, #agents, #events, or #results), so reload, direct links, and browser Back/Forward restore that view. Without a tab in the URL, Network Overview opens first. Submitting a run from the scenario editor, or continuing/retrying a saved batch, opens Exp Progress.

The Overview switch starts Off on every page load. Turn it On to fetch detailed topology and create the graph. While Off, Peer population and experiment metrics stay live without graph computation or rendering. Switching tabs does not enable it. Turning it Off releases graph data and selection; see topology controls.

Agent Status → Refresh queries registered Agents again, including offline ones, and updates their Agent/Peer state. The button shows progress and reports how many Agents refreshed; unavailable Agents are named while the existing list stays visible. This checks current status without restarting running Peers. More → Discover Agents also finds missing Swarm Agents and retries their registration.

Exp Progress → Refresh performs the same saved-result list refresh as Saved Results → Refresh, then updates both result views. Both buttons share the request and busy state. Refresh preserves filters, open result groups and the last loaded list if the request fails.

Agent Status presents the Agent + Peers, Agent only, and Peers only summaries above the Agent table, followed by Controller & Services. Tables scroll within their sections; measurement explanations can be expanded beside each section. Partial samples retain measured usage and show counts such as 46/47 containers. See resource scope and units for CPU normalization, coverage, and memory definitions.

• More → Current CSV downloads the latest resource values.
• Open Measurement & exports to use Start measurement → Stop. Select a completed interval, then use Actions → Interval CSV or Interval summary CSV. The same menu contains Delete interval.
• In Resource history, choose a period and use Export → History CSV or Summary CSV.

The measurement status remains visible while Measurement & exports is collapsed. The shared interval survives page reloads and Controller restarts, ends after at most 24 hours, and uses retained Prometheus samples. Refresh, measurement errors, and CSV feedback remain visible outside the collapsed controls.

Saved Results keeps Images and the ZIP download on each run, and Analyze batch mean on batch headers. Open that row's More menu for available Add runs, Retry, Continue remaining, or deletion actions. Expand a batch to access its individual runs. The same controls remain available on narrow screens without stacking every action above the results.

Experiments, background analysis, and live updates continue across tab changes. If Overview is On, leaving Network Overview stops layout animation; returning displays the latest topology while preserving filters, selection when the Peer still exists, and the Pause motion preference. The metrics row and its live values are shared across all tabs; switching tabs also preserves an expanded metric card.

Login and live updates

A session expiry opens Login required; log in again to return to the Dashboard. Closing or hiding the page stops its live connection, while accepted experiments and analysis continue on the Controller. Returning to a visible page synchronizes the current state. Reconnecting describes the browser connection, not the experiment outcome.

Use stream diagnosis for a connection that stays in recovery, authentication for session behavior, and Dashboard SSE for the wire format.

The Dashboard coalesces telemetry bursts into at most four renders per second and retains unchanged text, lists, and topology elements. Live metric updates preserve the open card and the horizontal scroll position while refreshing its values.

---

Korean translation: 한국어.