Core Concepts
The handful of ideas Zharp is built on: workspaces, monitors, regions, incidents, and the rules that decide who gets alerted.
Workspaces and roles
Everything in Zharp belongs to a workspace: monitors, incidents, status pages, integrations and the people who share them. You invite teammates from Team Members in the sidebar, and each member has one role — the one picked in the invite dialog, which can be changed later with Change role. The same four names are used everywhere.
| Role | What it allows |
|---|---|
| Owner | Full access, billing, and can delete the workspace. |
| Admin | Manage monitors, integrations and members. |
| Member | Create and manage monitors; acknowledge incidents. |
| Viewer | Read-only access to monitors and incidents. |
Monitors and their status
A monitor is one check that Zharp runs on a schedule: a website, an API, a port, a certificate, a database, a cron job and so on. Monitor Types lists them all. Every monitor is in one of these states:
| Status | Meaning |
|---|---|
| Pending | Created or resumed, and waiting for its first verdict. |
| Up | At least half of its regions report the check passing. |
| Down | More than half of its regions report the check failing. An incident is open. |
| Paused | You paused it. No checks run and no alerts are sent until you resume it. |
| Maintenance | A maintenance window covering it is in progress. Incidents and alerts are held back. |
To pause a monitor, click Pause on its page, or choose Pause from its menu in the monitor list. Resume puts it back to pending and checking starts again.
Checks, regions and going down
Zharp can check from four regions: Europe, North America, Asia and Australia. A new monitor uses all four. Every time the check is due, each selected region runs it and records its own result. Zharp then combines the latest result from each region, counting only results from the last five minutes:
- Down — more than half of the regions failed. With four regions that is three or four; with two regions, both.
- Up — half or more of the regions passed.
- Degraded — the monitor is up, but at least one region failed. The monitor page shows Failing in … under its status, but no incident opens and nobody is alerted: alerts follow the majority of regions. It clears once every region passes in the same round.
More regions make a monitor harder to fool. A single region going down on its own only marks the monitor degraded, not down, as long as most regions still reach it.
Confirmation and recovery periods
By default the change happens on the first combined result that says so. Two settings under Advanced settings make a monitor wait instead:
- Confirmation period — the regions must keep reporting the monitor down for this long before it goes down and an incident opens. Until then its status doesn't change, nobody is alerted, and Zharp checks again as the period ends. A passing result in between starts the period over.
- Recovery period — a down monitor must keep passing for this long before it goes back up and its incident closes. The default is 1 minute.
A failing region shows on the monitor page as soon as it is seen; a monitor waiting out its confirmation period isn't shown as degraded in the meantime.
Incidents
An incident is the record of one outage. Zharp opens it when a monitor goes down, and you can open one yourself with Report an incident on the Incidents page. An incident moves through these states:
- Ongoing — open, and nobody has responded yet. Alerts and escalation start here.
- Acknowledged — someone clicked Acknowledge. This tells the team the incident is being handled and stops escalation from paging further.
- Resolved — closed, either by someone clicking Resolve or automatically when the monitor passes its checks again.
If a monitor fails again with the same cause within 30 minutes of recovering, Zharp reopens the last incident and counts the recurrence, rather than opening a new one. See Incidents.
Alerts and escalation
Who is alerted
Each monitor chooses one of two ways to respond, under How should we escalate incidents? in its settings:
- Notify the on-duty member — the person on shift in your primary on-call schedule is alerted by E-mail or Push notification, whichever you tick. With no schedule, or no one on shift, the monitor's alert is all that goes out. Ticking Slack alert also posts to a Slack connection.
- Escalation policy — Zharp works through the steps of a policy you pick. A step can alert people, a group, an on-call schedule or an integration; show instructions; branch on the time or on metadata; or resolve the incident. Each step can wait a number of minutes before it runs, and the policy stops as soon as someone acknowledges the incident.
On-call schedules and escalation policies live under Team Members, on the On-Call Schedules and Escalation Policies tabs. See Escalation & On-call.
Alert channels
An alert channel is a place Zharp can post to: Slack, Telegram, Discord, Microsoft Teams, Google Chat, Mattermost, PagerDuty, Splunk, Zapier, Pushover, Pushbullet, a webhook, or the Zharp app for Android and iOS. You connect them under Integrations & API. See Alert Channels.
Separately from all of this, the workspace owner and everyone subscribed to a monitor get an email whenever it goes down or comes back up. You are subscribed to the monitors you create.
Maintenance and status pages
Maintenance windows
A maintenance window is a period when you expect monitors to fail — a deploy, a migration, a hardware change. Open Maintenance in the sidebar, click Schedule maintenance, set From and To, and select the monitors under Affected services. While the window is in progress those monitors show Maintenance, no incidents open and alerts are paused. When it ends, they are checked normally again.
This holds for every monitor type, log monitors included. An escalation already paging people about one of the monitors stops when the window starts, and the incident's timeline says so. Subscribers of the status pages showing those monitors are e-mailed when the window is scheduled and again when it starts.
Turn on Recurring maintenance and choose Every day or Every week to repeat the window.From and To then set the first occurrence; it repeats at the same time of day — every day, or on the same weekday every week — in your browser's timezone, following its clock changes. A daily window can last up to 24 hours and a weekly one up to 7 days. The list shows each recurring window's current or next occurrence.
A single monitor can also have its own recurring window, set in the Maintenance section of its form: a daily time range on the days you pick, in the timezone you choose. See Monitor Types. Those windows are managed from the monitor and do not appear in the sidebar list.
Status pages
A status page shows the state of the monitors you choose, so your users can see what is working without asking. Create one from Status Pages in the sidebar. See Status Pages.