> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/continuous-delivery/troubleshooting-and-resources/knowledge-base-article/cv-new-deployment-metrics.md).

# Continuous Verification with new deployment metrics

How Continuous Verification handles metrics that appear only on canary pods after a new deployment

In a canary deployment, Continuous Verification compares canary pods against primary pods to validate code changes. That comparison works when the same metrics exist on both canary and primary pods. When the canary introduces a metric that primary pods do not emit, you can still verify it with the AI Verify (v1) step by understanding how Harness treats missing primary (control) data.

***

## What you will learn from this topic

* How Continuous Verification behaves when a metric exists only on [canary pods](#default-behavior)
* How [Absolute Fail-Fast thresholds](#absolute-fail-fast-thresholds) can fail verification without primary data
* How to combine this with [Fail On No Analysis](#solution-with-fail-on-no-analysis) using a separate AI Verify (v1) step

***

## Before you begin

Confirm the following before you configure canary-only metrics in Continuous Verification:

* **Canary deployments:** Understand how Harness compares canary and primary instances. Go to [Create a Kubernetes canary deployment](/continuous-delivery/use-continuous-delivery/deploy-services-on-different-platforms/kubernetes/kubernetes-executions/create-a-kubernetes-canary-deployment.md) to review the strategy.
* **AI Verify (v1) step:** Know how to add and configure Continuous Verification. Go to [Verify deployments with the Verify step](/continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/verify-deployments.md) to configure health sources and thresholds.

***

## Problem statement

In a canary deployment strategy, the canary pod is compared against the primary pod (or pods) to verify the code changes. This works when the metrics used for comparison are present on both the primary and canary pods. If the canary pod introduces a new metric, you can still verify that metric with the AI Verify (v1) step by using the behaviors described below.

***

## Solution without Fail On No Analysis

### Default behavior

When a metric is emitted only by the canary pod, there are no pre-deployment pods for that metric, and all post-deployment pods for that metric are canary pods. Continuous Verification recognizes only canary pods and no primary pods. Canary pod data is collected as test data, but because there is no primary pod data (control data), the metric is not analyzed and the overall analysis result for that metric is **No Analysis**. By default, verification passes for this metric when no fail-fast threshold is breached.

![Default Continuous Verification result showing No Analysis for a canary-only metric and an overall pass](https://3694223630-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fy1JhZ4oKIppwY7d5AhPj%2Fuploads%2Fgit-blob-e56b6c6f2029e209c6310243becf0ce4138241f2%2Fcv-new-metric-default.png?alt=media)

### Absolute Fail-Fast thresholds

To ensure the canary pod is performing correctly for a canary-only metric, configure Absolute Fail-Fast thresholds for that metric. Verification fails if the threshold is breached, even when primary node data is absent.

![Continuous Verification failing because an Absolute Fail-Fast threshold was breached on a canary-only metric](https://3694223630-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fy1JhZ4oKIppwY7d5AhPj%2Fuploads%2Fgit-blob-4d4d933199a700833b60d36ca4fb84e9095d93c1%2Fcv-new-metric-absolute.png?alt=media)

***

## Solution with Fail On No Analysis

If **Fail On No Analysis** is enabled on the original AI Verify (v1) step, verification fails even when no threshold has been breached. Verify the new metric with a separate AI Verify (v1) step (serial or parallel to the original AI Verify (v1) step) where **Fail On No Analysis** is disabled.

![Pipeline layout using one AI Verify (v1) step with Fail On No Analysis and a second AI Verify (v1) step for the new metric](https://3694223630-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fy1JhZ4oKIppwY7d5AhPj%2Fuploads%2Fgit-blob-e5335207c92bc62fec8778d7927a69920373080e%2Fcv-new-metric-split.png?alt=media)

***

## Related concepts

* [Verify deployments with the Verify step](/continuous-delivery/use-continuous-delivery/verify-deployments/configure-cv/verify-deployments.md): Configure sensitivity, duration, and fail-fast options.
* [Create a Kubernetes canary deployment](/continuous-delivery/use-continuous-delivery/deploy-services-on-different-platforms/kubernetes/kubernetes-executions/create-a-kubernetes-canary-deployment.md): Set up the canary strategy that Continuous Verification analyzes.

{% @harness-feedback/feedback %}