Beta Beta pricing: Cleat is 75% off, 49 left at this price. See how it works 75% off during the beta. Details
Docker Swarm

Your cluster, watched.

Point Cleat at a Swarm manager and it polls the cluster over the SSH connection you already have. Six views over one cluster, and an alert engine that catches the failures docker service ls renders as perfectly healthy.

Cleat's Swarm pane over a three-node cluster: health tiles for services, tasks, nodes and updates, a service list with two failing services, and a critical alert in the corner.
No agent
Cleat shells docker over SSH to a manager you already have a connection profile for. Nothing to install on the cluster.
Read-only by default
Control is opt-in per cluster. Until you turn it on, Cleat can look and not touch.
Health at a glance
Services converged, tasks running, nodes up, updates in flight — with trend sparklines.
Every action audited
Who ran what, against which cluster, and when.
Why it exists

The deploy that looks fine.

A rolling update that fails and pauses leaves the service sitting at its old replica count. docker service ls renders that as 2/2 — indistinguishable from healthy. You find out when someone asks why the fix didn't ship.

The alerts inbox: services with no running replicas, a node down, two updates paused, and replicas not on the expected image — each with how long it has been firing.

Cleat ships enabled rules for no running replicas, replicas below desired, paused and rolled-back updates, rejected tasks, restart storms, image drift, unreachable managers and drained nodes. Edit them, change severities, or turn them off.

Nodes & tasks

Down to the failing task.

Roles, leadership and availability per node — and every task with its desired state, current state and the error Docker actually returned. Columns resize; double-click a row for the whole record, untruncated.

The nodes table: a manager marked leader and two workers, one of them down, with engine version, task count and committed CPU and memory. The tasks table filtered across a cluster, showing rejected tasks with the registry error that caused them.
Before it runs

Nothing happens by accident.

Cleat is pointed at production. Every cluster-changing action shows the verbatim docker command and what it will touch before it runs — and the ones that can't be undone by pressing the opposite button make you type the service name first.

A confirmation dialog for rolling back a service: the exact docker command it will run, what a rollback does, and a field requiring the service name to be typed. The audit tab listing control actions taken against the cluster.
Tune it

Rules you can read.

Every rule is a toggle, a severity and a scope — not a query language. Route them to a webhook, or run a local command when one fires.

The alert rules editor: seeded rules for replicas, rolling updates, rejected tasks, restart storms and image drift, each with a toggle and a severity. The events feed, showing what changed between cluster polls.
Hold · Fast.
Watch the cluster from the window you already have open.
Download free trial All features