Stoker

Alert channels

Events at or above the minimum severity are sent to every configured channel.

Minimum severity Quiet window (minutes) (the same alert is not re-sent within this window, stops a crash-looping service from emailing every restart cycle; 0 = send every time) Send a resolution notification when a check or agent recovers (paired with the down/offline alert, so on-call knows to stand down; off = recovery events still record to the log & dashboard, they just never dispatch) Webhook URL (POSTs JSON; payload includes a Slack/Discord-friendly "text" field. Leave empty to disable)

Email (SMTP)

SMTP host (empty = disabled)

Port

From

To

Username (optional)

Password (optional)

Use SSL/TLS

Disk space

Agents already report every fixed volume. These thresholds decide when a full disk becomes an alert, and they apply to every machine in the fleet.

Alert on low disk space

Warning at (% used)

and free space under (GB) (0 = ignore)

Critical at (% used)

and free space under (GB) (0 = ignore)

Both halves must be true. Percentage alone misfires at scale: 2% free on a 50 TB array is a terabyte, which is not an emergency, while 5% free on a 100 GB system disk is. The free-space figure is what makes one setting safe across volumes of wildly different sizes. Set it to 0 to go back to percentage only.

Remind me while it stays low

Ignore volumes smaller than (GB)

Never alert on these drives (comma separated, e.g. D:,E:)

Small volumes are skipped by default because EFI and recovery partitions sit near full by design and would alert on every machine you install. A volume has to climb clear of a threshold before it counts as recovered, so a disk hovering on the line does not alert and clear repeatedly. Going from warning to critical always notifies immediately, wherever the reminder schedule has got to.

HTTP uptime checks

Stoke and check your web applications via the hub.

Add

Agent API keys

Each agent presents a key in its X-Api-Key header. Key value is shown only once, at creation.

General

Updates

Maintenance windows

A window mutes alerts for the machines it covers, optionally holds their auto-restart, and afterwards checks that every service running when it opened is running again. Nothing stops being recorded: the event history and the results below are built from what happened while it was quiet.

Name These machines

Tags are the quick way to say "this client's servers". Add them on the dashboard, or ship them in the agent's config.

Repeat

Starts at

Lasts minutes

Time zone

Hold agent auto-restart while the window is open

Leave this on for anything that reboots or stops services. Turn it off when you only want the alerts quiet and Stoker should keep protecting the box.

Check that services come back, and alert if they do not

The reason to leave this on: it is what tells you at 03:20 that the box rebooted and SQL Agent never started, instead of the client telling you at 08:30.

Advanced

Wait before checking minutes

Mute disk alerts too

Disk alerts fire through a window by default: a volume filling up during a patch run is exactly when you want to know.

Never flag these services (comma separated)

For a service on these machines that legitimately stops itself after a reboot. Services set to delayed auto-start are already left out, so this is usually empty.

Results

What happened while each window was open: how long every machine was down, and whether the services running before it started are running again. Alerts are held during a window; this is the record of what the hub kept watching anyway.

License

Users

Admins can change anything. Read-only users can view the dashboard and events but not make changes or see settings.