One row per service per node, by service name, as each node last reported it. expected is 1 for a service the node's packages declare, and actual is 1 when the agent found it running (its system service active, or its port answering). For a proxy, fronts lists where it passes traffic on to.
Good to know
Yes-or-no fields come back as 1 or 0.
A node that stops checking in keeps its last report, so its services can look fine long after it went quiet. Compare checked_at with the time, or see the fleet map, which marks such nodes stale.
