Your runbooks, inside the guardrails.
Every team has the checks it runs by hand at 2 a.m. Custom operations let you write one down once, as a trigger, one fix Varqeni already knows, whether a person approves it, and when a person is paged, and run it the same way every time.
- Check the heartbeat age
- If over 60 s, read the connector state
- If paused, confirm the source is still writing
- Ask on-call to approve a resume
- Reads: heartbeat, connector, source
- May do: resume, with approval
- Never: drop slots, reset offsets
- Records: every step, sealed
Described, not coded
A short file says when it applies, on which hosts or workloads, and the one fix it uses.
Same guardrails
Read-only identities, a list of allowed actions, a named approval and the kill switch.
Same evidence
Every run is recorded with its sources and sealed like any investigation.
Tested before trusted
Dry-run it against your past incidents. It runs on its own only after its own fault trials pass.
Every custom operation has the same parts
- Trigger
The signal that starts it, confirmed so noise does not.
- Investigate
What to read, through read-only identities.
- Policy
Which action is allowed, at what risk, and whether it needs a person.
- Act and verify
Run the approved action once, then check the system is healthy.
- Escalate
Page a person with the evidence if it did not work, or roll back where that is possible.
“If disk use on a host crosses 85% because old rotated logs piled up, clear them, check the space came back, and only page the team if it did not.”
In the Fortis lab this runs as a runbook on the hosts it lists, with a person approving each run from the page.
Live in lab What it does today
- Runbooks as governed operations: a trigger, one fix Varqeni knows, verification, a page if it did not work
- Several steps with checks between them: wait, check it is still failing, run the fix, check it stays healthy, and page a person if any step fails
- One runbook shared by several pipelines or systems of the same kind
- A person approves each run, or it runs on its own after its own fault trials, within a rate limit
- Dry runs against your past incidents before a runbook is trusted, across every system it is shared with
Tell us the runbook you would automate first
Varqeni is in early access. Tell us about your pipelines and we will set up a trial around them.