When DevOps experts need to track down what causes errors in your app, it may not be easy to identify the cause. Using artificial intelligence, New Relic APM's error profiles automatically compare one set of events to another.
Each error profile provides visual details about significant differences in the frequency of different values for the events. For each attribute, the error profile includes:
- A pie chart showing how the error's attribute is distributed for values that deviate the most
- A table comparing the error attribute's distribution to that of non-erroring transactions
This helps you take more of the guesswork out of resolving your app errors. You can more easily determine if you can safely ignore the error, or if you should attempt to resolve the error with a new deployment, code edits, customer communications, or other actions.
Access to this feature depends on your subscription level.
To get a high-level overview of all your applications and services, use the entity explorer in New Relic One.
Error profile attribute examples
Error profiles appear as a separate tab on your New Relic APM Error analytics page.
- Error profiles feature in New Relic APM
An error profile is a collection of attributes with significantly different traits compared to non-errors. An attribute is "unusual" if a set of events represent what is normal (for example, errors compared to all traffic for a given time window), or differences between similar criteria (for example, two different hosts).
Errors may be related to events such as:
- Specific web transaction names or non-web transaction names, JVM thread names, etc.
- Unique types of error messages, classes, etc.
- Random customer interactions; for example, a particular error comes from a single customer's account, while normal traffic comes from a wide variety of accounts
- External call counts or duration
- Timing differences among hosts in your ecosystem, cluster agent IDs, etc.
- Other anomalies
Select error profile criteria
Based on criteria in your app's Error analytics page, New Relic analyzes and lists unusual trends by their significance. Your selected criteria includes:
- Time window
- Error analytics page filters
- Search criteria on the Error analytics page or the Error profiles tab
As you examine error profile results and want to dig deeper, add or change your app's error profile criteria. The Error profile tab refreshes to show the traits that distinguish the errors that match the updated criteria.
- Error profile criteria example
Your app's Error profile tab currently shows several error classes or messages. To filter to a specific error class or message, use any of these options:
- From the Error analytics page's time picker, change the time range. For example, change the default (30 minutes) to Last 24 hours, ending now.
- From the Error analytics page's filter [filter icon] section: Select Back to groupings list, then select other Error groups, Error attributes, or Custom attributes.
- From the Error analytics page's filter [filter icon] section: Click a specific item on the list to narrow the filter. For example, if several error messages are listed, click only the message you are interested in.
- From the Error profiles tab's search [search icon] box, type
message, or other search values.
Analyze error profile results
To examine details for the attribute results in your app's error profile:
- Go to rpm.newrelic.com/apm > (select an app) > Events > Error analytics.
- From the Error analytics page, select the Error profile tab.
- From the Error profile tab, review the list of error attributes that match the currently selected error profile criteria.
- To view a specific attribute's details, click it.
- To highlight specific error details, mouse over any pie chart segment or table row for the attribute.
- To investigate a specific attribute for your app's errors, type its name in the Error profiles tab's search window, or change the currently selected error profile criteria.
Compare values with large differences to identify the traits that distinguish the errors for an attribute. The comparative data in the error profile results and the error trace details can help you decide what steps to take for additional troubleshooting and resolving the error.
Error and non-error distribution
Depending on an error's attributes, sometimes the attribute is distributed differently for errors than for non-errors.
- Top deviating values
New Relic analyzes each attribute for your app's errors and compares the distribution for errors that match your criteria to transactions without errors. If the proportions between these errors are roughly the same compared to transactions without any errors, the attribute does not contain much useful information for debugging.
New Relic limits the error profile's pie chart and table for each attribute to show only the top deviating values. When proportions are roughly the same, New Relic does not include them in the error profile.
- Other category
After the top deviating values, the error profile groups the rest into an Other category. This helps you focus on the values that are different for these errors.
- No value category
If values are unusually present or are not present in the errors, you may see a No value category on the list of error profiles.
If you do not have any filters selected, the profile shows any distinguishing traits your errors exhibit in aggregate.
Example: A certain kind of transaction makes up 20% of all of your traffic and is responsible for 80% of your errors. The error profile will show the unexpected proportions in the
Adjust your error profile criteria to drill down even deeper into the profile results, so you can more effectively troubleshoot and resolve specific error events.