Exceptions Overview

​

How exceptions are generated

Metric run → FAIL
     │
     └── Execute Failed Rows Query
               │
               └── Each row → Exception Record
                             (asset, metric, status, owner, SLA, timestamp)

​

Prerequisites

1

Use a supported storage provider

Exception data is written to cloud storage. GCP, Azure, and AWS storage providers are supported. An administrator sets the provider in Organization → Platform → Storage.

2

Turn on Write Exception under Remediation

Open Organization → Remediation and turn on Write Exception in the Exception section. You can limit it to selected sources and set a row limit and exception categories. Write Exception must also be enabled on each metric (see the metric step below). It is required at both levels.

3

Define the schema under Remediation

In the Compute section of Organization → Remediation, enter the Schema that Prizm uses for exception processing.

4

Set up the Spark configuration

The Spark agent must be registered and deployed. Check it in Organization → Platform → Configuration → Agent Settings, where the spark agent should show a status of ACTIVE and an update status of DEPLOYED.

5

Use verified metrics with scoring enabled

The metric must be in VERIFIED status, and Scoring must be turned on for it. Exception writing must also be enabled at the metric level. In the metric’s Query Metric Configuration, this appears as the Export toggle with a row limit.

6

Have an alert and invalid records

An alert must be present for the exception to be fixed, and the metric run must return invalid records. The metric header shows the number of alerts and the valid and invalid record counts.

​

Where to find exceptions

​

Exception record statuses

​

SLA status

​

Enabling exception writing

1

Open the metric configuration

Open the metric’s detail page and click the pencil (edit) icon in Metric Detail.

2

Enable Write Exception

In the Configuration tab, find the Exception section and toggle Write Exception on.

3

Add a Failed Rows Query

In Query Metric Configuration, select the Failed Rows tab (next to Custom Query and Total Records) and write a SQL query that returns the individual rows contributing to a failure. This query runs automatically whenever the metric fails. The results become exception records.

-- Example: return rows with null order_id
SELECT order_id, customer_id, order_date, created_at
FROM orders
WHERE order_id IS NULL

4

Save

Click Save Changes. Exception writing takes effect on the next metric run.

​

Escalation

• Reports To: Each member has a Reports To manager, set in Access → Member. The Member hierarchy view (the second icon at the top right of the Access page) shows the tree with each member’s role and number of direct reports, and a search box to find a member.
• Role display names and owner mapping: In Access → Role, each role has a Display name and a Show in owner mapping switch (off by default). When it is on, the display name appears as a row on the asset’s Owners panel, and the user picker for that row lists only members who hold the role.
• Owner filter: The Exception dashboard in Analytics has a Member filter that shows the hierarchy, so you can view exceptions for a member or their reporting team. Exceptions without an owner appear under Unassigned.
• Notifications: Notifications are sent when an owner or status changes. See Exception notifications.

​

Comparison and distribution report templates

• Exception (Default): the standard exception report.
• Distribution: a distribution table of exceptions. The table uses Metric Criteria in place of the former rule column. Columns appear in this order: Critical, High, Medium, Low, followed by Critical %, High %, Medium %, and Low %.
• Comparison: a comparison-style report for exception results.