The alert inbox
The Alert inbox is where alerts are worked. It lists the alerts of the entity chosen in the header and everything below it, most severe first and then most recent, and opens the detail of an alert in a panel beside the list. This article explains the filters, how to read a row and the detail panel, and how root cause grouping works.
Filters
The inbox opens on the active alerts you have not seen yet, so what is left is what needs your eyes. When your installation does not track who has seen what, it opens on all active alerts.
| Filter | Values |
|---|---|
| Unseen by me | On or off |
| Severity | Critical, Warning, Info; several at once, or none for all |
| Search alerts | Words in the alert titles |
| Status | Active (open, acknowledged or silenced), Open, Acknowledged, Silenced, Resolved, All |
| Assignee | Anyone, Me, Unassigned, or a person with access to the entity |
| Category | All categories, Alerts, Security findings, Advisories; shown when your installation has the security or updates modules |
| Entity | Any entity you can see; it also changes the entity in the header |
| Resource | Set when you follow See its alert history from a resource page. Show every resource clears it |
Every filter is part of the page address, so you can bookmark a filtered inbox or send it to a colleague, and the browser's back button undoes a change. Reset filters goes back to the defaults. On a phone, the less used filters fold behind Filters, which shows how many of them are narrowing the list.
When nothing is left, the inbox says why: You have seen every active alert., No alert matches these filters. or No alerts. Everything is within its thresholds.
Reading a row
| Part | Shows |
|---|---|
| Rail and badge | The severity, as a colour and a shape |
| Dot and bold title | You have not opened this alert yet |
| Title | The check and, when it applies, the item it concerns, for example Last full backup age — Sales |
| Status | Acknowledged, Silenced or Resolved; open alerts show none |
| Second line | The resource, its entity, the current value with its threshold, and the Jira ticket when there is one |
| Right side | How long it has been going on, how many times it fired (×12), how many people have seen it, and the assignee |
Mark seen clears the dot without opening the alert. Seen is information for your team; it does not change the alert's status.
The detail panel
Select a row to open its detail beside the list, or full screen on a phone. The panel has its own address, /alerts/<id>, so you can share it. Opening it marks the alert as seen by you.
From top to bottom, the panel shows:
- The value and the rule it broke, for example critical when > 90%, with where that threshold is set: the baseline, an entity by name, or this resource.
- Resource, Entity, Check and Since, with the number of occurrences. Who acknowledged, silenced, was assigned or resolved it, and when.
- Suggested action from the check catalog, and the note that explains why the threshold is where it is.
- For a Cannot reach alert, Checks that cannot connect, each with its error and how many failures in a row.
- For a change, such as a failover, Changes: what the value was and what it became.
- Root cause and dependents, when the alert belongs to a group (see below).
- Seen by: who opened it and when.
- Timeline: opened, escalated or eased, acknowledged, assigned, silenced, notified (with the channels), a failed notification being retried, the Jira ticket and the resolution with its reason.
- The actions your role allows: Acknowledge, Silence, Assign, Resolve and Create Jira ticket. A viewer sees Your role on this entity lets you read this alert; operators and above act on it. Resolved alerts take no more actions.
Acknowledge and work an alert explains each action.
Root cause grouping
When one failure explains others, Excubius links them so you work the cause instead of a handful of unrelated alerts. Today this happens when Excubius cannot reach a resource: after three connection, sign-in or timeout failures in a row it opens one Cannot reach alert, and every other alert of that resource depends on it.
- The detail of a dependent alert shows Caused by with a link to the cause.
- The cause shows how many alerts depend on it, for example 2 alerts depend on this one, with links to each.
- Dependents are not notified while the cause is open, so your channel hears about the problem once.
Work the cause. When the resource answers again, the cause resolves on its own, and each dependent either resolves with its next check or carries on as an ordinary alert.