# Incidents and e-mail subscriptions

Receive by e-mail what matters to you, and only that.

## Alert or incident

An **alert** is a risk: something could go wrong. An **incident** is a fact: a node, a cluster, an array controller or pool has become unavailable, a source no longer answers, a backup job fails after succeeding before.

Incidents only come from live data, from a collector or a connector. An imported file does not open any.

## The steps

1. Open **Incidents**, under **Monitoring**. The red menu counter gives the number of open incidents.
2. Read the **Timeline**: one row per object, one bar per incident, over 24 hours, 7 days or 30 days.
3. Below the timeline, read what is open: the severity, the object, where it is, and since when. The **Notified** column says how many e-mails went out, and the ticket opened in your ITSM if there is one.
4. Click **My alert subscriptions**, at the top of the page. You also get there from **My notifications**, in your account menu: **Which alerts**, then **Choose**.
5. For each kind of alert, choose what you receive and from which severity, then save.
6. If needed, limit the perimeter to certain sources, datacenters or clusters.

> Everything stays recorded and visible on the Incidents page, whatever you choose: your subscriptions only decide what arrives by e-mail.

## The tabs of the page

| Tab | What it shows |
|---|---|
| **Incidents** | What is open, and what was resolved in the last 30 days |
| **Problems** | An incident that keeps coming back: the same object several times, or several nodes of one cluster |
| **Downtimes** | Your maintenance windows: the incident stays recorded, with no e-mail and no automatic analysis |
| **Monitored VMs** | The machines you want to follow like a node, on time slots if needed |
| **Notifications** | What Naoshift sent for the incidents of the period: each e-mail, per recipient, and the operations on tickets in your ITSM |

## Before maintenance

On an open incident, or from the **Downtimes** tab, click **Set a downtime**: a node, a cluster or a machine, a start, an end and a reason. Your on-call team is no longer woken up for planned work. This action is reserved for administrators and operators.

## Understand an incident

**Investigate** starts a root-cause analysis on the object: what changed around it, the events, and the logs when the source can read them. For a critical node or cluster incident, it starts by itself. You find all analyses under **Monitoring › Investigations**.

### I receive no incident e-mail

Check in **My notifications** that **Receive incident alerts** is on. This switch turns all these e-mails off at once, whatever the detailed settings. Check your spam folder too.

### A node stopped for maintenance triggered an alert

Set a downtime before the work: it covers a node, a cluster or a machine. On a cluster, it covers its nodes.

### How do I set a colleague's subscriptions?

An administrator sets those of any account, at the bottom of the same page. A backup team can thus receive only backup alerts.
