---
title: "Add a CloudWatch signal to deployment checks | Proofline Help"
description: "Open the environment’s deploy verification setup with the linked AWS account and region ready. Choose dimensions and a statistic that measure the service’s behavior."
url: https://proofline.sh/help/aws-add-deploy-signal/
---

[Help center](/help/)

# Add a CloudWatch signal to deployment checks

Open the environment’s deploy verification setup with the linked AWS account and region ready. Choose dimensions and a statistic that measure the service’s behavior.

## Configure the metric and its comparison rule

1.  Open the environment's **Setup → Deploy verification** page.
2.  Select **Add signal**.
3.  Check that **Provider** is **AWS CloudWatch**. The form starts there when the environment has an AWS account and no Google Cloud project; otherwise select it. If the form shows **Connect AWS**, the environment has no linked account. Link the account from section 2, then return.
4.  Select the linked **AWS account and region**. An account whose role cannot read CloudWatch metrics reads **cannot read metrics** in the list, and the form says why beneath the field and holds the save. When the account's read choices leave CloudWatch metrics out, select **Choose what Proofline reads**, include the metric read, and update the installed role. When its role denied the metric read, select **Update the AWS role**, which opens the connection; update the role from its installation files to grant the read in the region the form names, then select **Check access again** beneath the files or on the connection's **Status** tab. Proofline asks AWS for a refused metric read again only after that check, never on its own. Proofline checks each account in us-east-1 first, so when the form says us-east-1 denied the read of an account in another region, grant the read in both regions. A save refused for the same reason shows beside the field with the same step.
5.  Enter the signal's name, namespace, metric name and exact dimensions. Add its unit when needed.
6.  Set **Statistic**, **Worse when**, **Tolerance (%)** and **Minimum change** for that metric. **Data points per second** reads CloudWatch's `SampleCount`, the number of data points the metric published, not the requests or errors they count. For a total such as `HTTPCode_Target_5XX_Count`, choose **Sum per second**.
7.  For a signal that counts the service's requests with **Sum per second** or **Data points per second**, you may select **Counts the service's requests** and set **Traffic floor (requests per window)**, 100 unless changed. [Read a deploy's verification](/help/read-deploy-verification/#read-a-deploy) explains what a floor does.
8.  Select **Add signal**.

Leave **No data means zero** clear unless an absent value means zero for this counter. A metric that cannot be read leaves its check incomplete.

A traffic floor compares the signal's total in each window, so the signal must count every request of the service: a **Sum per second** of a request total, such as an Application Load Balancer's `RequestCount` for the service's target group. A signal of `HTTPCode_Target_5XX_Count` counts only the errors, and a **Data points per second** counts the data points the metric published, such as one a minute, however many requests each records. Either can count fewer than the floor while the service is busy.

## Related guides

-   [Turn on the production loop for AWS](/help/aws-production-loop/)
-   [Connect Amazon Web Services](/help/connect-aws/)
