Dashboard

Overview

Overview is where Flotilla opens, on an admin and on a host alike. It was called the dashboard until the fleet arrived; this page keeps its old file name, Dashboard, so existing links still work.

Overview is the fleet at a glance, and numbers only: is anything wrong, is anything ready to update, what does the fleet hold, and which Macs make it up — in that order, top to bottom. The per-Mac charts and tables it used to have moved to each Mac's own page under Hosts; see Each Mac's page below.

Needs attention

First, in a box of its own: what needs you. One notice at a time — a long list made Overview unusable. The notice's heading and detail are on the left; on the right, its fix where Flotilla has one (Start, Retry, Download Kernel), Dismiss where it is allowed, the ‹ 2 of 12 › arrows, and All, which opens Notifications. Errors come first, then the newest.

The box has a fixed height, so stepping through notices, or one resolving, never moves the page under you. When nothing needs you it says so in one line — "Nothing needs attention on 4 Macs" once hosts are paired — with All Notifications beside it.

Errors cannot be dismissed, only dealt with. Dismissing anything else hides it until it changes.

While container is being installed, or is missing, Overview's top line is the setup banner instead, the same one This Mac's page has. "Nothing needs attention" under a banner saying container isn't installed would contradict it, so the box only appears then if something else needs you too.

Get started

A card with the first things to do: Run a Container…, Pull Image…, Add Host… (on an admin) and Try a Suggested Stack…. It appears once container is usable, and stays until you close it with the X — having containers already does not hide it. Once closed it stays closed.

Updates

Shown only when there is something to update: hosts running an older Flotilla than This Mac, with Update 2 Hosts Now, and hosts that can upgrade container to the version Flotilla expects, with Upgrade 2 Hosts… — the count is the real one. One row each, however many hosts. An update is work to do, not a fault, so it lives here rather than in Needs attention — a routine rollout used to turn the menu-bar badge red.

Upgrading container asks first and says what it will stop: each host downloads Apple's installer, its Flotilla Helper installs it after checking it is Apple's, one host at a time, and container restarts on each. See Updates.

Across all hosts

Four tiles — Containers (running of total), Images, Volumes and Networks — each a link to its section. The numbers are This Mac's plus what each paired host last reported, so "running of 14 · 3 Macs" means across the fleet. Groups, DNS domains, machines and clusters are not summed here: they stay This Mac's.

Overview loads every list itself when it opens, rather than showing zeros on a fresh launch, and it asks the hosts alongside rather than first — a host slow to answer must not hold This Mac's numbers at zero.

Hosts

A table, one row per Mac: This Mac first, then every paired host, so thirty minis read as well as three. The heading says how many are connected (2 of 3 connected). The Host column has a coloured dot and the Mac's name, which opens that Mac's page under Hosts; the dot's state is connected, checking…, busy, not answering, container isn't installed or runtime unavailable.

The other columns are Tags, Model, CPU, Memory, Disk, macOS, Flotilla, container and Last Check-in. The Columns button at the table's top right (Show or hide columns) lists them with a checkbox each; every column but Host can be hidden, and the choice is kept per window across launches. See Views and filtering for how rows work. Memory and Disk are what the Mac has installed — 64 GB, 2 TB — not how much is in use: this is an overview. Last Check-in is kept through failures, so a silent host says since when. Up to ten rows show, then the table scrolls.

The version in the status bar

The Flotilla version sits in the status bar at the bottom left, so one glance at a host says whether its update has landed. On an admin Mac the version is a menu: it has Sparkle's Check for Updates…, the update channel, and the fleet's updates. On a host the menu says the admin updates it. See Menu bar and window.

It used to run Flotilla's own GitHub release lookup instead — a second update mechanism beside Sparkle that could disagree with it. That lookup is gone.

Each Mac's page

Hosts ▸ a Mac is the old dashboard's job, for every Mac. Its tabs are Overview · System · Flotilla · Settings · Updates · Activity, and its Overview has the Mac's own notices, its State, Hardware, Running here counts, Versions, the categories you have given it, and Usage now: CPU, memory used and disk free as it last reported them. See Hosts.

There are no live charts any more, on This Mac either. A host cannot send them, and keeping them on one Mac would have made This Mac the one page that differs.

The numbers, and why they differ from Activity Monitor

Two things here are easy to get wrong, and Flotilla got both wrong once before measuring properly.

Memory "used" is App Memory + Wired + Compressed. An earlier version summed active + inactive + wired + compressed and reported memory at 96–98% permanently. inactive_count is reclaimable file cache, and Activity Monitor excludes it — which is why the gauge always read full and was therefore worse than no gauge at all. The current figure is cross-checked against top at the same instant.

RAM is reported in binary units, disk in decimal. A 64 GiB Mac reads "64 GB" everywhere else in macOS, so reporting it as "68.72 GB" — which is what a decimal formatter does — invites the obvious question. Disk sizes stay decimal because that is the convention for storage. The two conventions genuinely differ by medium, and matching each is better than picking one and being wrong half the time.

Refresh cadence

Containers and their stats have two separate intervals — Refresh containers every and Refresh stats every in Settings — because sampling stats is more expensive than listing containers and tying them together would mean paying the higher cost for both. Machines, images, volumes and networks refresh every sixth container tick.

That is a deliberate compromise. Refreshing all five collections every tick would spawn several processes a second for state that barely changes; refreshing them only when you look at them meant the Activity feed and the sidebar counts were silently wrong until you visited the screen.

Paired hosts are asked for their status every 30 seconds while they answer, and less often while they do not, whether or not the window is open — so the menu bar knows when a host stops answering.

Activity

Activity has its own place in the sidebar, below Hosts with Logs. It is a feed of what changed, with Subject and Kind filters, Clear filters when you have set both and want out, and drill-down from an entry to the thing it happened to — the container itself, not just its section. Right-click adds Filter to that subject and a Copy menu. Each Mac's page has its own Activity tab. Jobs you have just started are in Tasks.

Two kinds of entry end up in it:

• Observed transitions — the poll loop comparing one refresh with the next.
• Performed actions — recorded by whatever performed them.

Both exist because a poll loop cannot see an action with no net state change. A restart of a running container ends with it running, so comparing refreshes finds nothing; a stop-then-start from the CLI produced no event at all. Start and stop each leave a lasting state, which is why only those two ever showed up before actions were recorded directly.