Notifications¶
A backup system you don't watch is a backup system that fails silently. Backupkrake sends push messages through ntfy and/or Pushover. Both are set up on the Notifications page.
Setting up ntfy¶
- Pick a topic. On the public
ntfy.shserver anyone who knows the topic can read it, so use something hard to guess, e.g.backupkrake-x7f3k2q9. - Subscribe to the topic in the ntfy app on your phone.
- On the Notifications page, enable ntfy, enter the server (
https://ntfy.shor your own instance) and the topic. - For a protected topic on your own server, open Authentication and enter an access token
(
tk_…), or a username and password. The token takes precedence. - Save, then Send test.
Priority on failure (default 4, high) is used for failures, overdue alerts, low free space and Send test. Warnings, including the fill-up forecast, are sent with priority 3, and "OK again" messages with 2.
Setting up Pushover¶
- Copy your user key from the Pushover dashboard (a group key works too).
- Create an application at pushover.net/apps/build and copy its API token.
- Enable Pushover, enter both, Save, then Send test.
Priority on failure (default 0, normal; 1 = high bypasses quiet hours) is used for the same events. Warnings and "OK again" messages are sent with -1 (no sound).
You can use ntfy and Pushover at the same time. Secrets you've saved are never shown again; type a new value to replace one, or tick remove saved value.
When you are notified¶
| Event | Notified | Setting |
|---|---|---|
| Host backup failed | always | |
| Backup looks incomplete (plausibility check) | always | |
| Content check failed | always | |
| Offsite backup or offsite restore test failed | by default | Offsite backup failed |
| Restore to a client failed | by default | Restore failed (dry runs only notify when they fail) |
| Run finished with warnings (unreadable or vanished files, post-command failed) | off by default | Finished with warnings |
| First good run after a failure, or after an overdue alert | by default | Backup works again |
| Host overdue | per overdue setting | |
| Backup storage low or filling up | per storage setting |
Not notified: cancelled runs, runs skipped outside a backup window or because a host was unreachable with retries enabled, and runs interrupted at the end of a window. They're normal operation. If they keep a host from ever finishing, the overdue alert catches it.
Overdue backups¶
Independently of individual runs, an hourly check alerts when a host has had no successful backup for
N days (default 2, 0 = off). Hosts can override it in their settings, e.g. 1 for a critical
server or 14 for a laptop.
- It reminds you once a day while the problem lasts, and sends Backups up to date again when it's resolved.
- Overdue hosts are marked on the dashboard.
- It catches what never produces a failed run: a schedule that was switched off, a host that's always outside its window, a scheduler that stopped, a server that was down.
The first check runs about two minutes after Backupkrake starts.
Backup storage running low¶
Every hour Backupkrake records how full /data is. From the growth over the last 30 days it estimates
when the disk will be full. The dashboard shows the estimate: full in about 48 days (+1.2 GiB/day).
A notification is sent when
- free space drops below 10 %, or
- the forecast says the disk is full within 30 days.
Both thresholds are configurable (0 = off). Reminders repeat once a day, and an OK again message follows when it clears. The forecast starts after two days of data.
To get space back: lower the retention of the hosts that use the most (see their disk usage card), exclude large files that change often, or delete old snapshots.
Heartbeat¶
Backupkrake can't tell you that it is itself down. For that, set a heartbeat URL, for example:
- a healthchecks.io check with a period of 1 hour and a grace time of 1 hour, or
- an Uptime Kuma push monitor with an interval of 3600 seconds.
Backupkrake fetches the URL every hour. When the pings stop (server off, container crashed, scheduler stuck), the external service alerts you.
Message contents¶
Messages include the run summary and the last lines of the log. With Backupkrake URL set (e.g.
http://nas.lan:8080), tapping a notification opens the run.
Delivery errors are logged and shown on the Notifications page (Last delivery). A broken notification service never affects backups.