Intermittent Service Requests
A degraded response time or increased failure rates for a service on SAP Commerce Cloud can indicate intermittent issues.
Site slowness or failure can cause intermittency, but the most common scenario occurs when you observe issues when a specific server receives a request. The quickest way to see this instance is to analyze the requests per service instances using the Service details page. You can also perform this analysis using the service flow or backtrace.
Service Instance Analysis
Narrow your investigation to instances for which the metrics show the poorest results. In these cases, visit the process detail page of a particular instance to analyze the metrics such as CPU, Memory, Threads, and DB Pool. For logs, use Kibana to further your investigation.
- From the sidemenu, select .
- Select service to analyze from the list of services.
- From the Requests panel, click View requests.
- Click the tabs that reveals the metric you want to analyze.
- Scroll down to the list of service instances.
The metrics that appear in the list of services depends on the currently active tab.
Service Flow Analysis
Service flow allows you to visualize the journey of a request in terms of infrastructure, starting from the service that you are analyzing. Since the first layer in our infrastructure is the web server, start the service flow analysis by filtering for Apache HTTP Server from the list of technologies in Services.
- From the sidemenu, select .
- To filter by Apache servers, check the Apache HTTP Server from the Technology panel on the left.
- Select a service that ends in 443.
- From the Understand dependencies panel, click View service flow.
The service flow helps to to discover when an intermittent failure takes place. For example, you could discover that an intermittent failure to redirect to the correct service only happens when the requests land on a particular backoffice pod. You would then look in the logs for more infomration, and then take the relevant corrective measures. For example, you could notice from the logs that backoffice Spring context failed to start only on a specific pod.
The service flow can also help to detect other possible problems related to intermittency, such as increased slowness on specific instances or failed requests that you can then attribute to load balancing.
Service Backtrace Analysis
Analyze backtrace is similar to service flow, although it works in reverse in the sense that it traces the requests back to the end user. Starting the analysis of the backtrace from the database end provides the most comprehensive view of the request since it includes both the Tomcat and Apache layers.
- From the sidemenu, select .
- To filter by Apache servers, check the Apache HTTP Server from the Technology panel on the left.
- Select a service that ends in 443.
- From the Understand dependencies panel, click Analyze backtrace.
Starting you backtrace analysis from the database allows you to investigate any of the services for more details, including an overview of all requests involving the service, and then broken down by instances.
For Tomcat services, you can analyze the stack trace leading to the request. For example, if you click a storefront service, and then e above and then enable a stacktrace of all requests made to the product page on ProductPage.
Depending on the selected service, some dimmed options may become available. For example, the Client IPs tab becomes available when you select Calls from unknown hosts, pages and devices, which correspond to the end user requests.
Even though Dynatrace masks the last digit of the Client IP (replaced with 0), you can still discover valuable insight regarding intermittent issues at the end user level by revealing the number of failed requests. The same is true for the Instances tab when it comes to intermittent issues on the pod level.


