Skip to content

Work Units (Jobs & Pipelines)

Work Units is LakeSentry’s unified view of Databricks jobs and Delta Live Tables (DLT) pipelines — the scheduled and triggered workloads that drive most of your compute spend. Each work unit tracks a single job or pipeline across all of its runs, giving you a cost history and performance baseline.

The main page shows all jobs and pipelines across your connected workspaces:

ColumnWhat it shows
NameJob or pipeline name (with creator shown underneath)
TypeJob or pipeline
SignificanceCategory badge (Top Spender, Growing, or Long Running)
TrendCost sparkline for the selected time range
CostTotal cost for the selected time range
Cost/RunAverage cost per run
RunsTotal number of runs (all time)
Avg DurationMean execution time
SuccessSuccess rate as a percentage
Last RunWhen the work unit last executed

Sort by name, cost, significance score, or last run time. Common sorting strategies:

  • Cost descending — Find the most expensive work units
  • Significance descending — Focus on what matters most
  • Last run descending — See the most recently active work units

Filters available:

FilterOptions
WorkspaceSpecific workspace
TypeJobs, Pipelines, or All Types
SearchFilter by name or external ID

Every work unit receives a significance score based on cost impact, execution frequency, and historical reliability. Badges help you prioritize:

BadgeConditionMeaning
Top SpenderSignificance score >= 45High cost impact — always worth monitoring
GrowingCost change > 15%Cost is trending upward compared to the prior period
Long RunningAvg. duration > 30 minExecution time is notably long

Click any work unit to see its dedicated detail page.

A bar chart showing daily cost over time. Each bar represents one day’s total cost for the work unit. The chart includes:

  • Trend area — A shaded area overlay showing the cost direction over the period
  • Average line — The mean daily cost across the visible range

This visualization makes cost spikes immediately visible. A day that stands well above the average is likely worth investigating.

A detailed table of all runs in the selected time range:

ColumnWhat it shows
Run IDDatabricks run identifier
ComputeWhich cluster or warehouse executed the run
StatusRun result state (e.g. SUCCEEDED, FAILED, ERROR, CANCELLED, TIMEDOUT)
Start timeWhen the run started
DurationWall-clock execution time
CostTotal cost for this run (DBU + cloud)

Aggregate metrics for the work unit over the selected period:

MetricWhat it measures
Total costSum of all run costs (with DBU/cloud breakdown)
Total runsNumber of runs, with recent runs in the last 7 days
Avg. durationMean execution time
Success ratePercentage of runs that completed successfully

The detail page includes a cost breakdown by Databricks SKU (billing category), showing how the work unit’s cost distributes across different SKUs such as Jobs Compute, Jobs Light, or All-Purpose Compute. Each SKU shows its cost and proportion of the total.

The detail page includes a panel with key metadata:

  • External ID — The Databricks job or pipeline ID (with a link to open in Databricks)
  • Workspace — Which workspace the work unit belongs to
  • First Seen — When the work unit was first observed
  • Last Run — When the work unit last executed
  • Significance — The significance score (if available)

Shows upstream and downstream table dependencies for the work unit, helping you understand data lineage. Upstream tables are those read by the work unit, and downstream tables are those written to, along with their consumers.

While both appear as work units, there are some differences in how they are tracked:

AspectJobsDLT pipelines
Run granularityPer job runPer pipeline update
Cost sourcePer-run billing via job_run_id linkageDaily cluster usage aggregates
Cluster infoJob cluster or existing clusterPipeline cluster (auto-provisioned)