Skip to main content

Enterprise Troubleshooting

This page covers the most common issues encountered during Pixee Enterprise deployment and operation. Each entry describes the symptom, likely cause, and resolution steps. Issues are organized by deployment phase: installation, scanner connectivity, LLM configuration, authentication, and day-two operations. If your issue is not listed here, contact Pixee solutions engineering.

Installation Issues

SymptomLikely CauseResolution
Embedded cluster install hangs or times outInsufficient system resourcesVerify minimum requirements: 8 vCPU, 32 GB RAM, 100 GB+ SSD/NVMe with under 10 ms write latency. Check disk I/O performance.
Helm install fails with image pull errorsContainer registry access not configuredVerify image pull secrets are created in the Pixee namespace and the cluster can reach the container registry.
Air-gapped install fails to startMissing container images in local registryVerify all required images have been transferred to the local registry and are tagged correctly. Compare against the image manifest.
KOTS admin console is inaccessiblePort forwarding or firewall rulesVerify port 30000 is accessible from your browser. Check firewall rules and any network policies that may block the admin console port.
Install completes but UI is unreachableTLS configuration mismatchVerify the domain and TLS configuration in the admin console matches your DNS and certificate setup. Self-signed certificates require browser trust or exception.

Minimum Requirements Quick Check: Before troubleshooting any installation issue, confirm: 8 vCPU, 32 GB RAM, 100 GB+ SSD/NVMe disk, ports 443 and 30000 open (embedded cluster) or port 443 open (Helm). Most installation failures trace back to insufficient resources or network access.

For detailed setup procedures, see Embedded Cluster, Helm / BYO Kubernetes, or Air-Gapped Deployment.

Scanner Connectivity

SymptomLikely CauseResolution
Scanner findings not appearing in PixeeSARIF file format mismatch or upload path errorVerify the scanner output conforms to the SARIF specification and the upload endpoint is correctly configured.
Partial findings ingestionScanner output exceeds size limits or contains unsupported rule IDsCheck the scanner output for unsupported finding types. Verify the scanner is producing complete SARIF output.
Duplicate findings after re-scanDeduplication key mismatch between scan runsVerify the scanner configuration produces consistent finding identifiers across runs.
Findings appear but no fixes generatedScanner findings are in an unsupported language or CWE categoryCheck the Languages Overview for supported language and finding type coverage.

For scanner-specific setup guides, see Integrations Overview. For custom scanner configuration, see the Universal SARIF guide.

LLM Configuration

SymptomLikely CauseResolution
Fixes not generatingLLM endpoint unreachable or credentials invalidVerify the LLM endpoint URL is correct and the API key is valid. Run the preflight check from the admin console to test connectivity.
Slow fix generationLLM provider rate limits or network latencyCheck your provider's rate limit status. Verify the network path between the Pixee cluster and the LLM endpoint has acceptable latency.
Air-gapped LLM not respondingPrivate LLM endpoint misconfiguredVerify the endpoint is accessible from within the cluster network namespace. Check proxy configuration if the endpoint requires proxy access.
Triage decisions seem inaccurateLLM model capability insufficientThe quality of triage and remediation depends on the model at the configured endpoint. Lower-capability models produce lower accuracy. Review the model assigned to the Reasoning tier.
Preflight check fails during installationEndpoint URL, API key, or network path errorRead the preflight error message carefully -- it specifies whether the issue is authentication, connectivity, or configuration.

For LLM provider setup details, see Bring Your Own Model.

Authentication and Access

SymptomLikely CauseResolution
SSO login failsOIDC configuration mismatchVerify the issuer URL, client ID, and callback URL match between Pixee and your identity provider configuration.
Users can log in but cannot see repositoriesSCM permissions not syncedRe-sync repository access from the SCM platform. Verify the service account or GitHub App has access to the expected repositories.
Role assignment not workingRBAC misconfigurationCheck user role assignments in the admin console. Verify the identity provider is sending the expected role claims.
SSO redirect loopCallback URL mismatchVerify the callback URL configured in Pixee matches the redirect URI registered with your identity provider exactly, including protocol and trailing slashes.

For SSO configuration details, see Security Architecture.

Day-Two Operations

SymptomLikely CauseResolution
High disk usageLog retention or metrics storage growing unboundedAdjust log retention settings in the admin console. Configure metric pruning intervals. Review support bundle log size limits.
Grafana dashboards not loadingObservability subchart disabled or misconfiguredVerify the Grafana subchart is enabled. If using BYO observability, verify telemetry is flowing to your platform.
Upgrade failsHelm values conflict with new versionReview release notes for breaking changes before upgrading. Back up your values file before running the upgrade.
Support bundle is too largeDefault log collection scope too broadConfigure support bundle log size and age limits before generating. Narrow the collection scope to the relevant time window.
Analysis queue growing without processingLLM endpoint degraded or rate limitedCheck LLM endpoint health and rate limit status. Review the error rate in the observability dashboard.

For monitoring configuration, see Observability.

Getting Help

If your issue is not covered on this page:

  1. Generate a support bundle from the KOTS admin console (embedded cluster) or using the support bundle tool (Helm). Configure log size and age limits before generation to control bundle size.

  2. Contact Pixee solutions engineering with:

    • Deployment model (embedded cluster, Helm, air-gapped)
    • Pixee version (visible in admin console or Helm release)
    • Symptom description and when it started
    • Support bundle (attached)
  3. For critical issues affecting production analysis, escalate through your enterprise support channel for priority response.