Code Quality Testing code-quality-testing

Learn how code quality testing of pipelines works and how it can improve the quality of your deployments.

Introduction introduction

During pipeline execution, a number of metrics are captured and compared to either the key performance indicators (KPIs) defined by the business owner or to standards set by Adobe Managed Services.

These are reported using three-tiered rating system.

Three-Tiered Ratings three-tiered-ratings

There are three gates in the pipeline:

For each of these gates, there is a three-tiered structure for issues identified by the gate.

Code Quality Testing code-quality-testing-step

This step evaluates the quality of your application code, which is the main purpose of a code quality only pipeline, and is executed immediately following the build step in all non-production and production pipelines. Please refer to the document Configuring Non-Production Pipelines to learn more.

Code quality testing scans the source code to ensure that it meets certain quality criteria. This is implemented by a combination of SonarQube analysis, content package-level examination using OakPAL, and dispatcher validation using the Dispatcher Optimization Tool.

There are over 100 rules combining generic Java rules and AEM-specific rules. Some of the AEM-specific rules are created based on best practices from AEM Engineering and are referred to as Custom Code Quality Rules.

The results of code quality testing are delivered as rating as summarized in this table.

Dealing with False Positives dealing-with-false-positives

The quality scanning process is not perfect and will sometimes incorrectly identify issues which are not actually problematic. This is referred to as a false positive.

In these cases, the source code can be annotated with the standard Java @SuppressWarnings annotation specifying the rule ID as the annotation attribute. For example, one common false positive is that the SonarQube rule to detect hardcoded passwords can be aggressive about how a hardcoded password is identified.

The following code is fairly common in an AEM project, which has code to connect to some external service.

@Property(label = "Service Password")
private static final String PROP_SERVICE_PASSWORD = "password";

SonarQube will then raise a blocker vulnerability. But after reviewing the code, you recognize that this is not a vulnerability and can annotate the code with the appropriate rule ID.

@SuppressWarnings("squid:S2068")
@Property(label = "Service Password")
private static final String PROP_SERVICE_PASSWORD = "password";

However, if the code was actually this:

@Property(label = "Service Password", value = "mysecretpassword")
private static final String PROP_SERVICE_PASSWORD = "password";

Then the correct solution is to remove the hardcoded password.

Security Testing security-testing

Cloud Manager runs the existing AEM Security Heath Checks on the staging environment following deployment and reports the status through the UI. The results are aggregated from all AEM instances in the environment.

These same health checks can be executed at any time through the Web Console or the Operations Dashboard.

If any of the instances reports a failure for a given health check, the entire environment fails that health check. As with code quality and performance testing, these health checks are organized into categories and reported using the three-tiered gating system. The only distinction is that there is no threshold in the case of security testing. All the health checks are pass or fail.

The following table lists the health checks.

Performance Testing performance-testing

AEM Sites aem-sites

Cloud Manager executes performance testing for AEM Sites programs. The performance test is run for approximately 30 mins by spinning up virtual users (containers) that simulate actual users to access pages in staging environments to simulate traffic. These pages are found using a crawler.

Virtual Users virtual-users

The number of virtual users or containers that are spun up by Cloud Manager is driven by the KPIs (response time and pageviews/min) defined by the user with the Business Owner role while creating or editing the program. Based on the KPIs defined, up to 10 containers that simulate actual users will be spun up. The pages that are selected for testing are split and assigned to each virtual user.

Crawler crawler

Prior to the start of the 30 minute test period, Cloud Manager will crawl the staging environment using a set of one or more seed URLs configured by the Customer Success Engineer. Starting from these URLs, the HTML of each page is inspected and links are traversed in a breadth-first fashion.

Page Sets for Testing page-sets

Pages are selected by three page sets. Cloud Manager uses the access logs from the AEM instances across production and staging environments to determine the following buckets.

Distribution of Traffic Across Page Sets Selected distribution-of-traffic

You can choose anywhere from one to all three sets on the Testing tab of your pipeline configuration. The distribution of traffic is based on the number of sets selected. That is, if all three are selected, 33% of the total page views are put toward each set. If two are selected, 50% goes to each set. If one is selected, 100% of the traffic goes to that set.

Let us consider this example.

Over the 30 minute test period:

Testing and Reporting testing-reporting

Cloud Manager executes performance testing for AEM Sites programs by requesting pages as an unauthenticated user by default on the staging publish server for a 30 minute test period. It measures the virtual user-generated metrics (response time, error rate, views per minute, etc.) for each page as well as various system-level metrics (CPU, memory, networking data) for all instances.

The following table summarizes the performance test matrix using the three-tiered gating system.

Refer to the section Authenticated Performance Testing for more details on using basic authentication for performance testing for Sites and Assets.

Authenticated Performance Testing authenticated-performance-testing

If necessary, AMS customers with authenticated sites can specify a username and password which Cloud Manager will use to access the website during sites performance testing.

The username and password are specified as pipeline variables with the names CM_PERF_TEST_BASIC_USERNAME and CM_PERF_TEST_BASIC_PASSWORD.

The username should be stored in a string variable and the password should be stored in a secretString variable. If both of these are specified, every request from the performance test crawler and the test virtual users will contain these credentials as HTTP Basic authentication.

To set these variables using the Cloud Manager CLI, run:

$ aio cloudmanager:set-pipeline-variables <pipeline id> --variable CM_PERF_TEST_BASIC_USERNAME <username> --secret CM_PERF_TEST_BASIC_PASSWORD <password>

Please refer to the Patch User Pipeline Variables API documentation to learn how to use the API.

AEM Assets aem-assets

Cloud Manager executes performance testing for AEM Assets programs by uploading assets repeatedly for a 30 minute test period.

Onboarding Requirement onboarding-requirement

For Assets performance testing, your Customer Success Engineer will create a cloudmanager user and password during the onboarding of the author to staging environment. The performance test steps require a user called cloudmanager and the associated password set up by your CSE. This should neither be removed from the author instance nor its permissions changed. Doing so will likely fail the Assets performance testing.

Images and Assets for Testing assets-for-testing

Customers can upload their own assets for testing. This can be done from the Pipeline Setup or Edit screen. Common image formats such as JPEG, PNG, GIF and BMP are supported along with Photoshop, Illustrator and Postscript files.

If no images are uploaded, Cloud Manager will use a default image and PDF documents for testing.

Distribution of Assets for Testing distribution-of-assets

The distribution of how many assets of each type are uploaded per minute is set in the Pipeline Setup or Edit screen.

For example, if a 70/30 split is used, and there are 10 assets uploaded per minute, 7 images and 3 documents will be uploaded per minute.

Testing and Reporting testing-and-reporting

Cloud Manager will create a folder on the author instance using the username and password as setup by the CSE. Assets are then uploaded to the folder using an open-source library. The tests run by the Assets testing step are written using an open source library. Both processing time for each asset as well as various system-level metrics are measured across the 30-minute testing duration. This feature can upload both images and PDF documents.

Performance Testing Results Graphs performance-testing-results-graphs

A number of metrics are available in the Performance Test dialog

The metric panels can be expanded to display a graph, provide a link to a download, or both.

This functionality is available for the following metrics.

Content Package Scanning Optimization content-package-scanning-optimization

As part of the quality analysis process, Cloud Manager performs analysis of the content packages produced by the Maven build. Cloud Manager offers optimizations to accelerate this process, which are effective when certain packaging constraints are observed. Most significant is the optimization performed for projects that output a single content package, generally referred to as an “all” package, which contains a number of other content packages produced by the build, which are marked as skipped. When Cloud Manager detects this scenario, rather than unpack the “all” package, the individual content packages are scanned directly and sorted based on dependencies. For example, consider the following build output.

If the only items inside myco-all-1.0.0-SNAPSHOT.zip are the two skipped content packages, then the two embedded packages will be scanned in lieu of the “all” content package.

For projects that produce dozens of embedded packages, this optimization has been shown to save upwards of 10 minutes per pipeline execution.

A special case can occur when the “all” content package contains a combination of skipped content packages and OSGi bundles. For example, if myco-all-1.0.0-SNAPSHOT.zip contained the two embedded packages previously mentioned as well as one or more OSGi bundles, then a new, minimal content package is constructed with only the OSGi bundles. This package is always named cloudmanager-synthetic-jar-package and the contained bundles are placed in /apps/cloudmanager-synthetic-installer/install.