Image Enrichment Tool

Overview

When Klaudia detects an image-related issue during Root Cause Analysis (RCA) - such as ImagePullBackOff, ErrImagePull, or an invalid image tag - she doesn't stop at identifying the problem. Klaudia goes a step further by enriching her investigation with broader context from across your cluster, helping her suggest more accurate and reliable remediations.

This capability is powered by the Image Enrichment Tool, which Klaudia invokes automatically when image issues are identified in an RCA.

What Klaudia Does

When an image-related failure is detected, Klaudia automatically performs four types of analysis:

1. Scope Identification

Klaudia finds all pods across the cluster that are running - or have previously pulled - the same container image, and surfaces their current pod phases (Running, Pending, Failed, etc.). This helps answer: "Is this a widespread issue or isolated to one deployment?"

2. Failure Pattern Detection

Klaudia extracts and categorizes container failure reasons - ImagePullBackOff, ErrImagePull, RegistryUnavailable, InvalidImageName, and others - to distinguish between different failure types and pinpoint the root cause. It also identifies which workloads (Deployment, StatefulSet, etc.) own each failing pod, helping scope the impact accurately.

3. Tag Comparison Analysis

Klaudia identifies pods running different tags of the same image. This helps answer questions like: "Does tag v1.0.1 work when v1.0.0 fails?" pointing to a version-specific issue and informing rollback or roll-forward decisions.

4. Registry-Wide Health Stats

Klaudia analyzes all pods pulling from the same registry and calculates the ratio of failing vs. healthy pods. This answers: "Is this a registry-wide outage or a problem specific to this image?"

• Registry-wide failures → likely an infrastructure or authentication issue
• Image-specific failures → likely an image build or configuration issue

Why This Matters

Without this context, a possible fix might be switching to a different image tag - without knowing whether that image is actually valid or accessible in your environment. Image Enrichment makes Klaudia's suggestions more grounded in the reality of your cluster, reducing the chance of recommending an image that would fail to pull for the same underlying reason.

Permissions & RBAC

Image Enrichment relies on Klaudia's ability to query resource information across your cluster. This capability is subject to your account's RBAC permissions.

• If the user triggering the investigation does not have permission to view cross-namespace or cross-service resource data, Klaudia will work with the information available within the permitted scope.
• Users with broader view permissions will benefit from richer enrichment results and more accurate suggestions.

No additional configuration is required. Klaudia respects your existing RBAC policies automatically.

Example Scenario

A deployment fails with ImagePullBackOff on image my-registry.io/app:v2.3.1.

With Image Enrichment, Klaudia will:

• Check scope - confirm whether any other pods in the cluster are using the same image and whether they are healthy or also failing.
• Detect failure patterns - identify whether the failure reason is consistent across pods or varies, and which workload types are affected.
• Compare tags - check whether other tags of my-registry.io/app are running successfully, which may indicate a version-specific issue with v2.3.1.
• Assess registry health - determine whether other services pulling from my-registry.io are also failing, distinguishing a registry-wide outage from an image-specific problem.

Using this evidence, Klaudia can give a precise recommendation - for example, pointing to the specific pull secret that should be referenced, confirming the tag doesn't exist, or suggesting a known-good alternative already running in the cluster.