---
title: Alert rules, incidents and uptime for hosts
description: A full disk or a quiet agent used to be something you noticed on a chart. Every host now has Alerting, Incidents and Uptime SLA tabs, like a monitor: three…
canonical: https://watchfor.io/changelog/2026-09-15-hosts-alerting-incidents-sla
---

[All updates](/changelog/page/3#2026-09-15-hosts-alerting-incidents-sla) ✨ New Sep 15, 2026 · v 7.8.5

# Alert rules, incidents and uptime for hosts

A full disk or a quiet agent used to be something you noticed on a chart. Every host now has Alerting, Incidents and Uptime SLA tabs, like a monitor: three rules come switched on (agent silent, disk almost full, memory pressure) and a dozen more are one click away — CPU, load per core, I/O wait, swap, inodes, zombie or blocked processes, a watched process that is not running, a reboot. They notify through the same channels, on-call rotations and maintenance windows as everything else. The host page also gains Settings with notes, self-explaining cards, and a monitor-style header.

Agent Silent, Disk Almost Full and Memory Pressure come switched on.

## Three rules from the start

Every host comes with these switched on, so a new server is covered before you touch anything:

- Agent Silent: the agent stopped sending, because the server is down, unreachable or the agent stopped.

- Disk Almost Full: the fullest filesystem reaches 80% (warning) or 90% (critical).

- Memory Pressure: memory in use stays above 85% (warning) or 95% (critical).

## One click for more

High CPU, load per core, I/O wait, CPU steal, swap, inodes, blocked or zombie processes, network errors, a watched process that is not running, and a reboot. Rules with two levels have a slider for each, and each level can be [switched on or off on its own](/changelog/2026-09-15-alert-levels-switch-separately).

## Incidents and uptime

A firing rule opens an incident, exactly as for a monitor. The Incidents tab lists them and Uptime SLA reports on them. Uptime counts confirmed critical incidents only, so a silent agent counts as downtime and a busy CPU on its own does not.

## An example

Log rotation fails on web-2 and the root filesystem starts to fill. At 80% a warning lands in your Slack channel. At 90% a critical incident opens and pages whoever is on call. Once space is freed, the incident resolves by itself and the Incidents tab keeps the record.

Rules, thresholds and durations are in the [Hosts docs](/docs/hosts#alerting).

Since then: hosts start with your [default contact group](/changelog/2026-09-16-host-cards-alerting-defaults), and they can be picked in [maintenance windows](/changelog/2026-09-23-hosts-maintenance-and-limits), with alerts that name the disk or process.

Share this update

---

Canonical page: https://watchfor.io/changelog/2026-09-15-hosts-alerting-incidents-sla · Site guide: https://watchfor.io/llms.txt