Skip to main content

Troubleshooting

Common Issues​

"Re-run your install path" below means rerunning notebooks/install.py from the top (notebook path, recommended) or ./scripts/deploy.sh --update (local terminal path). Use whichever path you originally installed with — don't mix them for the same app instance.

SymptomCauseFix
App shows blank pagefrontend/dist/ missing (gitignored, not synced)Re-run your install path
Could not import module "backend.main"Source files missing on workspaceRe-run your install path (full sync)
No dependencies file foundpyproject.toml / uv.lock not on workspaceRe-run your install path
"Failed to list spaces"Lakebase not attachedAttach a postgres resource in Apps UI, or re-run your install path
Catalog 'X' is not accessibleWrong catalog or missing permissionsVerify the catalog widget value (notebook) or GENIE_CATALOG (terminal); databricks catalogs list --profile <profile>
Invalid SQL warehouse resourceWarehouse doesn't exist or no CAN_USEdatabricks warehouses list --profile <profile>
Maximum number of appsWorkspace hit the 300-app limitDelete unused apps
Unresolved __GSO_*__ placeholdersThe installer couldn't patch app.yamlEnsure the catalog is set; check installer output for warnings
GSO job creation fails during deployNotebook path: Jobs API error or wheel build failure. Terminal path: bundle deploy failed (CLI version, auth, build)Notebook: check the install cell output. Terminal: check databricks bundle deploy -t app output; ensure CLI >= 0.297.2 and pip install build
Notebook upload fails (RESOURCE_DOES_NOT_EXIST)Upload path not writable by deployerCheck workspace-level permissions on the upload path

Permission Errors​

SymptomCauseFix
"You need CAN_EDIT or CAN_MANAGE permission" on optimize triggerUser lacks permission on the Genie AgentShare the agent with the user (CAN_EDIT or CAN_MANAGE)
"The service principal does not have CAN_MANAGE"SP not shared on the Genie AgentShare the agent with the app's SP (CAN_MANAGE)
"OBO token lacks genie scope, retrying with service principal" (in logs)User token missing dashboards.genie scopeThis is handled automatically via SP fallback — no action needed unless SP also fails
Optimization job fails with catalog/schema access errorsSP lacks UC permissions on referenced dataGrant SELECT on referenced schemas to the SP
"Permission denied" on scanUser lacks access to the Genie AgentShare the agent with the user

Lakebase Issues​

SymptomCauseFix
"Failed to list spaces" on first loadLakebase not attachedRe-run your install path to auto-attach the postgres resource
Connection timeouts after ~1 hourCredential refresh failedCheck logs for generate_database_credential errors
Tables not created on startupSP lacks CONNECT or CREATE ON DATABASERe-run your install path to re-create the SP role and grants
Scan results not persistingLakebase write failedCheck logs for Failed to persist scan result
Agent sessions lost on restartLakebase not configuredWithout Lakebase, sessions use in-memory storage (ephemeral)

GSO / Auto-Optimize Issues​

SymptomCauseFix
"GSO not configured" in health checkGSO_JOB_ID or GSO_CATALOG not setRe-run your install path (patches app.yaml with the resolved job ID and catalog)
Optimization job never startsJob doesn't exist or SP can't run itCheck job exists in workspace; verify SP has CAN_MANAGE on job
Job stuck in QUEUEDNo available cluster or warehouseCheck cluster policies and warehouse availability
Baseline evaluation failsBenchmark questions reference inaccessible tablesGrant SP SELECT on all referenced schemas
Run ends SKIPPED with INSUFFICIENT_VALID_BENCHMARKSFewer than 15 valid benchmarks remained after QC and the bounded top-up budgetReview the top-up attempt telemetry and rejected findings, then add or repair benchmark questions before rerunning
Run ends FAILED with EVAL_INVALIDNative Eval-Run API returned an unusable result after retriesCheck the Eval-Run status in the run page; verify SP access to referenced tables
Patches generated but accuracy doesn't improveOptimization strategy exhaustedRun may reach STALLED status — review suggestions for manual improvements
__GSO_*__ values in running appThe installer didn't patch app.yaml before deployVerify the catalog value and re-run your install path

Debug Commands​

# View app logs
databricks apps logs <app-name> --profile <profile>

# Check app status
databricks apps get <app-name> --profile <profile>

# List workspace files to verify sync
databricks workspace list /Workspace/Users/<email>/<app-name>/backend --profile <profile>

# Check GSO job status
databricks jobs get <job-id> --profile <profile>

# List GSO job runs
databricks jobs list-runs --job-id <job-id> --profile <profile>

# Check SP identity
databricks apps get <app-name> --profile <profile> | grep service_principal

MLflow Tracing​

note

MLFLOW_EXPERIMENT_ID is workspace-specific. The app validates it at startup and silently disables tracing if the experiment doesn't exist. To enable tracing, create an MLflow experiment and update the value in app.yaml before deploying.