DEMO Live LakeSentry demo — no sign-up required. Try it now
For Platform Engineering Safe Databricks optimization

Get detailed observability into Databricks resources. Apply fixes safely.

Cluster lifecycle, idle warehouses, retry storms: surfaced as insights with evidence and ranked by impact. Real performance and cost optimization opportunities.

Start free

Read-only service principal. Unity Catalog compatible. Connect in minutes.

app.lakesentry.io/insights
Active Insights
23
Potential Savings
$12,450/mo
Needs Attention
7
4 critical, 3 high
Anomaly Critical
Cost spike detected
etl_daily_ingest
$2,100/mo 2 hours ago
Cost Multiplier 3.2x
Z-Score 4.1
Baseline Cost $1,240
Current Cost $3,968

Sample data. Your insights will reflect your environment.

You can see waste. Acting on it is the hard part.

A query is hitting a 256-GB cluster that runs all day, and the team that owns it is not on call. The fix is obvious: drop a node, change auto-termination, move the workload. The consequences of doing it wrong are also obvious. So the cluster keeps running.

Most platform teams have plenty of visibility. What they don’t have is a safe path from insight to action that survives the next on-call rotation.

  • Optimization scripts rot the moment whoever wrote them changes teams.
  • Custom watchers catch the anomaly you wrote them for, and miss everything else.
  • Audit trails for cost actions live in Confluence pages no one reads.
  • Business demands lower DBU spend without service degradation. No margin for error.

Visibility first, then a plan you can trust.

What LakeSentry surfaces today for platform engineers, and how each finding turns into a plan.

Cross-workspace visibility

Normalized cost and usage across every workspace. Cluster, job, and SQL warehouse health in one place, ranked by impact. No SQL stitching across regions or metastores.

Insights with evidence

Each recommendation ships with the data it ran on — z-scores, baselines, the deviation from history. Read the evidence, decide whether it applies. No black-box scoring.

Performance insights

Detailed insights and anomaly detection for all types of Databricks resources and workloads — clusters, SQL warehouses, pipelines, jobs, apps, endpoints, Lakebase databases, and more.

How LakeSentry stays safe.

What makes platform engineering teams comfortable connecting a vendor to production.

  • Read-only service principal. LakeSentry never asks for write access.
  • Approve and dismiss are admin-only RBAC. Every approval and every dismissal is captured in the audit log.
  • Outbound webhooks today cover scheduled reports (daily, weekly, monthly).
  • Nothing runs against your environment. Every plan is applied by your own team, in Databricks.

How the same fixes happen today — and with LakeSentry.

Manual reviews, custom scripts, or a vendor surface. The trade-off is time and trust.

Manual reviews Custom scripts LakeSentry
Coverage When you remember to look Whatever you scripted All workspaces, daily by default, hourly available
Where insights live Tickets and tribal knowledge Application logs Ranked queue with evidence per insight
Path to the fix Manual in the Databricks UI Whatever you wrote Guided by detailed action plans
Audit Confluence pages Application logs Findings, approvals and applied-plan history in the audit log
Trust posture Trust the reviewer Trust your own code Read-only default

Questions platform engineers ask.

What permissions does LakeSentry need?
Read-only. A service principal with access to the system tables is all LakeSentry needs, and it never requests write permissions. Unity Catalog is required.
How do I apply a recommendation?
Every recommendation ships as a guided action plan: what to change, the value it has today, and the effect to expect. You apply it in Databricks yourself — LakeSentry does not make changes in your environment.
How do I undo a change I applied?
Plans are generated with the current value of what they would change recorded inline — the prior auto-termination value, the previous worker count, whether the cluster was running. Undoing means applying that recorded value back in Databricks.
Do you integrate with our incident workflow?
Outbound webhooks today cover scheduled reports (daily digest, weekly, monthly), signed with HMAC-SHA256. Wire them into Slack, PagerDuty, or your own incident tooling. Per-action and per-insight webhooks are on the roadmap, not yet shipped.
Can I write custom rules?
The recommendation engine is built-in. Custom attribution rules, which decide how cost rolls up to teams, are fully yours to define in the Mappings tab and the rule editor.

Read-only access. Full cross-workspace visibility.

Free with unlimited workspaces. Unity Catalog compatible.

Start free