Skip to content
TechFabric

Build to compound

See what production is doing, and be allowed to act

Signals, severity from policy, and the authority to intervene. So the person who notices is the person who can fix it.

The alert fired. Then what.

Monitoring answers whether something broke. The expensive questions are the three it does not.

Nobody knows if it matters

A job failed. Whether that breach is urgent, and to whom, is worked out by whoever happens to know what that job feeds.

The person watching cannot act

They can see it and not stop it. The person who can stop it is not looking. The gap between those two is where the money goes.

The bill has no cause

Spend doubled last quarter. Which jobs did it, and whether any of them produced anything anybody used, is a forensic exercise.

Observation was never the hard part. Deciding what a signal means and who is allowed to respond to it is, and that is a governance problem wearing an operations hat.

The shift

Three ways to watch a live estate

Only one lets somebody do something about what they see.

Platform monitoring

It ran, or it did not

Job runs
Metric emitted
Alert fires
Triage by hand
Guess at impact

A dashboard somebody built

Accurate, and unread

Metrics collected
Dashboard drawn
Nobody watches
Found out later
Explained afterwards

Governed mission control

Scored, and actionable

Signal
Severity by policy
Intervention offered
Approved by someone allowed
Recorded

What gets stood up

The signals, the judgement and the authority, as one system rather than three tools.

Workload signals

Job state, latency, cost and data quality collected as events rather than scraped from a dashboard.

Severity from policy

Whether a breach matters is evaluated against rules somebody wrote, not inferred by whoever is on call at the time.

Interventions with authority

Stopping a workload is an action with a policy in front of it, so the person who sees the problem can be the person who ends it.

Cost attributed

Spend traced to the jobs and teams that caused it, so a growing bill has a name rather than a mystery.

Budgets that act

A budget that stops a workload, rather than one that emails about it after the quarter closed.

The response history

What fired, who acted, under which policy. Reading the estate's past is a query rather than an archaeology project.

Policy

Decides severity, rather than whoever is on call

Built for estates we were answerable for

Radar came out of running Databricks workloads where somebody was going to ask us what happened and why. It is free and open source and it runs in your workspace.

FAQ

Production visibility: common questions

We have monitoring already. What does this add?

The two things monitoring does not do: decide whether a breach matters, and let the person who saw it act. Those are policy questions rather than metric questions, which is why a monitoring tool does not answer them.

Will it actually stop a runaway job?

If you give it the authority to. The design deliberately makes that your decision rather than the tool's, because a system that can stop production without anybody agreeing to it is its own incident.

Does it work on estates we did not build?

Yes, and that is most of them. It reads what Databricks already emits rather than requiring the workloads to be instrumented for it.