How often can checks run?

Every 15 minutes, every 30 minutes, hourly or daily.

Will I get a notification every time the check fails?

No. There is one finding per device and problem while it stays open, without repeat notifications. It closes itself when a later check passes.

Where can notifications go?

Email over SMTP, any webhook, and DingTalk, Feishu or WeCom group bots. Channels are added under Settings → Notifications and selected per rule.

Can a check run my own script?

Yes. A custom-command rule runs a command on the device and treats a non-zero exit code, or output that contains or does not contain some text, as a finding.

Is the automatic fix safe?

The fix is a batch task you wrote and saved. It runs unattended, so any command that needs approval is skipped and noted rather than executed, and you can require approval before the fix starts at all.

Use cases · Monitoring

Scheduled health checks that can also fix what they find

Thirty shop PCs, a few servers and some Raspberry Pi boxes do not justify a Prometheus stack, but you would still like to know when a disk is nearly full or the POS service has died before a customer tells you. Octling’s inspections run simple checks on a timer on every connected device, notify you once per problem, and can start a fix on their own.

What you can check

Disk space: free space below a threshold on the system disk, all partitions or one path.
Missing process: a named process is not running.
Sustained high load: CPU above a level for a number of minutes.
Offline too long: no heartbeat from the device for more than a set number of hours.
Custom command: run any command on the device and treat a non-zero exit code, or output that contains (or lacks) some text, as a problem. This covers certificate expiry, a backup file that did not appear, a pending reboot, RAID state and whatever else you can test from a shell.
The client itself: outdated version, resident service not running, token about to expire.

Setting one up

Under Devices → Inspections, create a rule and pick the check and its condition.
Choose the scope: all devices, or particular groups and tags.
Choose how often: every 15 minutes, every 30 minutes, hourly or daily.
Choose what happens on a hit: record only, start a saved batch task, or restart the resident client. You can require approval before the fix runs.
Pick notification channels: email over SMTP, any webhook, or a DingTalk, Feishu or WeCom group bot.

One finding per problem

A device with a full disk produces one open finding, not a notification every 15 minutes. The finding stays open while the check keeps failing and closes itself when a later check passes. You can also mark a finding handled or ignore it.

Letting it fix things

The fix is a batch task you saved earlier, for example “clear old logs and package caches, do not touch anything under /var/lib”. When a rule hits, that task runs on the affected device. It follows the same approval rules as any unattended task: a command that needs approval is skipped and reported, not run. Start with “record only” until you trust the check, then attach the fix.

What it is not

This is not a metrics system. There are no graphs, no history of CPU or memory over time, and the shortest interval is 15 minutes, so it will not page you within seconds of an outage. If you need dashboards and second-level alerting, keep your monitoring stack and use Octling for the part where something has to log in and fix the machine. Checks also only run on devices that are connected; a switched-off PC shows up through the offline rule, not the disk rule.

Questions

How often can checks run?: Every 15 minutes, every 30 minutes, hourly or daily.
Will I get a notification every time the check fails?: No. There is one finding per device and problem while it stays open, without repeat notifications. It closes itself when a later check passes.
Where can notifications go?: Email over SMTP, any webhook, and DingTalk, Feishu or WeCom group bots. Channels are added under Settings → Notifications and selected per rule.
Can a check run my own script?: Yes. A custom-command rule runs a command on the device and treats a non-zero exit code, or output that contains or does not contain some text, as a finding.
Is the automatic fix safe?: The fix is a batch task you wrote and saved. It runs unattended, so any command that needs approval is skipped and noted rather than executed, and you can require approval before the fix starts at all.