twentyhq/twenty · error · Error

Failed to create searchVector GIN index(es) for workspace ${

Error message

Failed to create searchVector GIN index(es) for workspace ${workspaceId}

What it means

Thrown by the 2.20 searchVector GIN index reconciliation command during its applyBackfill phase. For each application that needs a missing GIN index, the command calls validateBuildAndRunWorkspaceMigration with index-creation operations. If the pipeline returns status 'fail', the full failure report is logged to logger.error (JSON-serialized), then a simpler error is thrown. The pipeline validates the index operations, builds migration steps, and executes them — failure means the index metadata could not be created.

Source

Thrown at packages/twenty-server/src/database/commands/upgrade-version-command/2-20/2-20-workspace-command-1783529458169-reconcile-search-vector-gin-index-universal-identifier.command.ts:256

                flatEntityToDelete: [],
                flatEntityToUpdate: [],
              },
            },
            workspaceId,
            applicationUniversalIdentifier,
          },
        );

      if (validateAndBuildResult.status === 'fail') {
        this.logger.error(
          `Failed to create searchVector GIN index(es) for application ${applicationUniversalIdentifier}:\n${JSON.stringify(
            validateAndBuildResult,
            null,
            2,
          )}`,
        );

        throw new Error(
          `Failed to create searchVector GIN index(es) for workspace ${workspaceId}`,
        );
      }
    }
  }

  private async flushIndexCacheAndBumpMetadataVersion(
    workspaceId: string,
  ): Promise<void> {
    const indexRelatedMetadataNames = [
      'index',
      ...getMetadataRelatedMetadataNames('index'),
      ...getMetadataSerializedRelationNames('index'),
    ] as const;
    const cacheKeysToFlush = [
      ...new Set(indexRelatedMetadataNames.map(getMetadataFlatEntityMapsKey)),
    ];

View on GitHub (pinned to 1f5dd2bbd2)

Solutions

Read the logger.error output — it contains JSON.stringify(validateAndBuildResult) with the exact failure reasons in the report
Flush and recompute the workspace cache, then retry: the command is idempotent (already-backfilled indexes are skipped)
If the report shows a universal identifier collision, verify the re-own phase completed: check indexMetadataRepository for legacy vs deterministic universalIdentifier values
Run with --dryRun to see which indexes the command intends to create without committing

Example fix

// before: throws on first application failure, aborting all remaining applications
if (validateAndBuildResult.status === 'fail') {
  throw new Error(`Failed to create searchVector GIN index(es) for workspace ${workspaceId}`);
}

// after: collect failures, continue, then report all at once
const failures = [];
// ... in the loop:
if (validateAndBuildResult.status === 'fail') {
  failures.push({ applicationUniversalIdentifier, validateAndBuildResult });
  continue;
}
// ... after the loop:
if (failures.length > 0) {
  throw new Error(`Failed for ${failures.length} application(s): ${JSON.stringify(failures)}`);
}

Defensive patterns

Strategy: retry

Validate before calling

// Before running, flush the workspace cache to ensure the builder sees fresh state
await workspaceCacheService.invalidateAndRecompute(workspaceId, [
  'flatIndexMaps',
  'flatObjectMetadataMaps',
  'flatFieldMetadataMaps',
  'flatApplicationMaps',
]);

// Verify no legacy universal identifiers remain that would collide
const legacyCount = await indexMetadataRepository.count({
  where: { workspaceId, /* legacy identifier pattern */ },
});

if (legacyCount > 0) {
  console.warn(`${legacyCount} legacy index rows may collide — re-own phase must succeed first`);
}

Try / catch

// Retry once after flushing cache, since validate-build-run failures are often cache-staleness related
try {
  await command.runOnWorkspace({ workspaceId, options: { dryRun: false } });
} catch (error) {
  if (error.message.includes('Failed to create searchVector GIN index')) {
    await workspaceCacheService.flush(workspaceId, ['flatIndexMaps']);
    await command.runOnWorkspace({ workspaceId, options: { dryRun: false } });
  } else {
    throw error;
  }
}

Prevention

Run upgrade commands with --dryRun first to validate the operation shapes before committing
Ensure no concurrent workspace metadata modifications are running during the upgrade
Flush the relevant workspace cache keys before retrying a failed upgrade command
Check the logger.error JSON output for the exact failure report before retrying — it distinguishes validation failures from DB errors

When it happens

Trigger: validateBuildAndRunWorkspaceMigration returns { status: 'fail', report: ... } when creating searchVector GIN index flat entities. Caused by: a duplicate universal identifier collision (legacy index survived re-own), a referenced object/field not existing in workspace metadata, stale workspace cache feeding invalid state to the builder, or a DB-level error during migration execution.

Common situations: The preceding re-own transaction (applyReOwn) partially failed or rolled back, leaving legacy index rows that collide with the deterministic identifiers; workspace cache is stale after a prior partial upgrade; concurrent workspace metadata modifications during the upgrade.

Related errors

AI-assisted analysis of twentyhq/twenty@1f5dd2bbd2 (2026-08-12). Data as JSON: /api/errors/a93a65f1ee8f9c8a. Report an issue: GitHub.