LakeSentry Common Issues Troubleshooting

Solutions to the most frequently encountered issues with LakeSentry. For Direct Connection data pipeline questions, see Data Freshness & Pipeline Status. For External Connector deployments, see Collector Issues.

Quick reference

Symptom	Likely cause	Jump to
Can’t log in or access the app	Authentication or invitation issue	Login and access problems
Dashboard shows no data	Connector not set up, sync not completed, or collector not running for External Connector deployments	Missing data
Costs don’t match Databricks console	Different time ranges, cost model, or attribution scope	Cost discrepancies
Metrics appear stale or outdated	Pipeline lag, Data Sync schedule issue, or collector schedule issue for External Connector deployments	Stale metrics
Some workspaces or resources missing	Region connector or permission gap	Incomplete coverage
Insights not appearing	Detection thresholds or insufficient history	Missing insights

Login and access problems

”Invitation expired” or no access

If you can’t log in or see a “no access” message:

1. Check your invitation — Invitations expire after 7 days. Ask your organization admin to resend the invitation from Settings > Access.
2. Verify the email address — You must log in with the exact email address the invitation was sent to. Check for typos or alias differences.
3. Check your role — If you can log in but can’t access certain features, your role may not have the required permissions. See Roles & Permissions for the permission matrix.

SSO or OAuth login failures

If authentication redirects fail or loop:

• Clear browser cookies for app.lakesentry.io and try again.
• Try an incognito/private window to rule out cached credentials.
• Check your identity provider — If your organization uses SSO, verify your account is active in the IdP and assigned to the LakeSentry application.
• Check the SSO configuration — Admins can find common causes (domain restrictions, pending access requests, redirect URI mismatches) in the SSO troubleshooting table.

Missing data

No data at all after setup

If the dashboard is completely empty after connecting your Databricks account:

1. Check validation and sync results — Ask an admin to open Settings → Connector. Confirm credentials validate, then inspect Data Sync run history and table-level errors. A successful connection test alone does not populate dashboards.
2. Check region connectors — Each region needs its own connector. If no region connectors are set up, there’s no data pipeline. See Region Connectors.
3. Check extraction status — Direct Connection uses the Data Sync panel in Settings > Connector. If you use an External Connector, verify that the collector has run at least once.
4. Check processing and filters — Inspect Ledger Load for active or failed jobs. On the dashboard, select a range with known usage and clear unintended filters. Your assigned access scope can also restrict results.

Tip

The most common reason for missing data after initial setup is that validation or the first sync has not completed. Follow the Quick Start Guide to ensure all setup steps are completed.

Some data types missing

If dashboards show cost data but certain features or resource types are empty:

Missing data	Likely cause	Fix
SQL warehouse data	Warehouse system tables not accessible	Grant SELECT on system.compute.warehouse_events to the service principal
Job/pipeline data	Missing system.lakeflow.* permissions	Grant access per Account & Connector Setup
MLflow data	MLflow data may not be available for your tenant	Ask LakeSentry support whether MLflow data is available
Model serving data	Serving system tables not accessible or unavailable	Grant access to serving system tables
Storage data	Storage system tables not accessible or unavailable	Grant access to storage system tables

Check the connector’s table coverage list to see which tables LakeSentry is successfully extracting.

Cost discrepancies

LakeSentry costs differ from Databricks console

Cost differences between LakeSentry and the Databricks billing console are common and usually explainable:

Time range alignment

• LakeSentry and Databricks may use different time zone conventions. LakeSentry uses UTC for all cost calculations. Verify both tools are set to the same date range.
• Source publication, the sync schedule, and processing all affect Data Freshness. Today’s costs may be partial, and late corrections can change past totals.

Cost model differences

• By default, LakeSentry uses Databricks list prices for Databricks costs. If your organization has negotiated pricing, the numbers will differ unless you’ve configured price overrides.
• LakeSentry separates Databricks costs from cloud infrastructure estimates where available. Compare Databricks Only mode with the equivalent billing scope before adding cloud costs; serverless and Lakebase charges do not imply a separate cloud cost component.

Scope differences

• LakeSentry shows costs only for workspaces and regions with active connectors. If some workspaces aren’t connected, their costs won’t appear.
• Attribution groupings (by team, project, or tag) may slice costs differently than the Databricks console’s per-workspace view.

If a difference remains after you align the time range, scope, freshness, and pricing, record both totals and their source views, then contact support.

Attribution shows “unattributed” costs

If a significant portion of costs appear as “Unattributed”:

• New resources — Resources without tag or ownership data default to unattributed. Configure Attribution Rules to assign them.
• Shared compute — Interactive clusters shared across teams may lack clear ownership signals. Use tag-based or pattern-based attribution rules.
• System overhead — Some Databricks costs (like platform fees or photon charges) don’t map to specific user workloads.

See Cost Attribution & Confidence Tiers for how attribution works and how to improve coverage.

Stale metrics

Dashboard data appears outdated

If metrics haven’t updated recently:

1. Check pipeline status — Go to Settings > Connector and look at the Last synced time of each connector. If it is older than the configured schedule plus expected Databricks latency, there may be an extraction or pipeline issue.
2. Check table and processing results — Expand Data Sync table details and history. Resolve failed tables, then inspect Ledger Load for processing errors.
3. Check for Databricks-side delays — Some system tables have inherent latency. See Data Freshness for how source delays affect each data type.

Metrics show zeros or flat lines

If metrics suddenly drop to zero:

• Extraction stopped running — Direct Connection may be paused or failing. For External Connector deployments, the Databricks job may have been disabled, deleted, or failed.
• Permissions revoked — The service principal may have lost access. Re-grant permissions per Account & Connector Setup.
• Region connector disconnected — Check connector status. If it shows “Error,” investigate the error message.

Incomplete coverage

Some workspaces are missing

If certain Databricks workspaces don’t appear in LakeSentry:

• Check region mapping — Workspaces in regions without a region connector aren’t monitored. Add a region connector for each active region. See Region Connectors.
• Check workspace access — The service principal must have workspace-level access in each workspace it should monitor. Account-level access alone isn’t sufficient for all system tables.

Historical data is missing

LakeSentry only shows data that was available to extract when the connector first synced and afterward. It cannot recover data beyond Databricks system-table retention.

• Billing data — Databricks system tables retain billing history. LakeSentry can extract historical billing data on the first run, covering up to the connector’s lookback setting (default 180 days, subject to plan limits).
• Compute and workload data — These tables have shorter retention. LakeSentry captures what’s available at first sync and tracks incrementally after that.

Tip

Connect LakeSentry as early as possible to maximize historical data capture. The longer you wait, the less historical compute and workload data will be available.

Missing insights

No anomalies or waste findings

If the Insights page is empty despite having cost data:

• Insufficient history — Evidence requirements vary by detector. Check the relevant Anomaly Detection or Insights & Actions guidance rather than applying one baseline rule to every finding.
• Below minimum thresholds — Each detector has its own eligibility and significance rules; low spend or too few observations may not produce a finding.
• No waste conditions detected — Waste detection looks for specific patterns (idle clusters, overprovisioned resources). If your environment is well-optimized, there may genuinely be nothing to flag.
• Auto-dismiss rules — Check if auto-dismiss rules are filtering out insights. Auto-dismiss rules are configured at the system level and can filter insights based on criteria like savings thresholds, age, or severity.

Insights aren’t updating

• LakeSentry checks for findings on a schedule and after each data processing run. A check updates an open finding instead of creating a new one. If extraction or processing is delayed, the evidence stays unchanged.
• Check the Data Freshness page for pipeline status.

Getting help

If the issue persists after the relevant sections above, send your LakeSentry support contact:

1. The affected page, tenant URL, workspace/resource identifiers, and exact error text.
2. The selected date range, filters, and cost mode; include both compared totals for a cost discrepancy.
3. The last successful sync and failure timestamps, including timezone.
4. Data Sync table errors and the relevant Ledger Load run/log details, or Databricks collector job run ID for External Connector deployments.
5. Recent credential, permission, network, or configuration changes, and the steps already tried.

Remove tokens, connection strings, OAuth secrets, and personal API keys from shared logs or screenshots.

Next steps

• Collector Issues — Troubleshooting External Connector deployments
• Data Freshness & Pipeline Status — Understanding data latency and pipeline health
• FAQ — Answers to frequently asked questions