Service detection and naming
Dynatrace automatically detects and names your applications' server-side services based on some basic properties of your application deployment and configuration. In case the default naming scheme does not cover your particular needs, you can customize the naming of your services.
As service landscapes can become quite complex, Dynatrace automatically categorizes services based on their dependencies to other entities like services or applications.
Dynatrace automatically groups related processes into process groups. When Dynatrace detects multiple process groups, it assumes that the process groups represent separate applications, or at least separate logical parts of a single application. Process groups therefore represent boundaries for the services they contain.
When Dynatrace detects the same service in multiple processes within the same process group, it represents the service as a single service running on multiple processes or hosts. Conversely, if Dynatrace detects a seemingly identical service in multiple process groups, it represents the separate service instances as separate services, even though the services appear to be the same. For this reason it sometimes makes sense to customize how Dynatrace detects process groups.
Web request services
Web request services serve the web applications that you deploy either via a web server (for example, Apache, IIS, or NGINX) or within web containers (for example, Java, .NET, Node.js, or PHP). Dynatrace considers three discrete concepts when identifying and naming web services: Web server name, Context root, and Web application ID.
There are three different terms—Virtual host, Server block, and Site—that represent similar concepts across different technologies.
- Virtual domain name hosting combines web requests from multiple hosts, domains, and IP addresses into a single configuration on a web server. For example, Apache HTTP Server allows you to configure
www.dynatrace.plall within the same virtual host.
- NGINX uses server block concept. In case of a server block, one needs to configure a
- Site is a concept established in Microsoft IIS.
In Kubernetes-based environments, Apache and NGINX web servers are often configured only as
localhost or just an IP address. In these cases, Dynatrace automatically uses the auto-detected base pod name as the web server name to provide more a meaningful representation out of the box.
Within any given web container you may have multiple applications in multiple directories. For example
/admin leads to the admin application while
/shop leads to the online store. In the Java world this is called a context root. Microsoft IIS refers to this concept as a Virtual directory. Most web servers don't even include this as a concept.
Some technologies allow you to assign specific names to your deployed web applications.
- For Java servlets, this is done by defining a display-name in the
- For Spring boot applications, you have to define the
spring.application.name, which may be in the
application.propertiesfile or the
- For Java technologies that don't allow for application naming (and to provide default application naming), you can define the Java command line property
-Ddynatrace.application.id=<name>. This won't override the naming options mentioned above—it serves as a default for instances when no application name is otherwise available.
- For Node.js, you can define a name in the
- For other technologies, or to provide a default name, you can use the environment variable
DT_APPLICATIONID=<name>. This won't override the naming options mentioned above—it serves as a default for instances when no application ID is available.
You can find these attributes in the Properties and tags of the service overview page.
Dynatrace picks up some or all of these properties and then creates a unique service based on them. When Dynatrace finds a web application ID, it uses the ID as the default service name. In other cases, web request services are named based on the web server name + the context root. This means that if you give an IIS site a proper name or define a name for your web application in
package.json, Dynatrace will pick up the name you specify.
Web services are defined by Web Services Description Language (WSDL), which is part of your deployment. WSDL defines services, how services are called, and service names. Dynatrace picks up service names along with
targetNamespace values and combines these values to uniquely identify each service.
Dynatrace detects web services based on technology-specific frameworks. For details on the web-service frameworks that are monitored by Dynatrace out-of-the-box, see supported web service frameworks for Java and .NET.
Sometimes it's technically not possible to easily detect a web service name. In such cases Dynatrace uses the web service endpoint rather than the name.
When Dynatrace detects that your application makes database requests, it identifies the name of the database or schema, the database vendor, and the IP address/port of the database. It uses this information to define a unique monitored database service and, where possible, detect on which process the database service runs.
For the full list of database services that are supported by Dynatrace, see supported database services.
The hostname you use to connect to Amazon RDS must be identical to the actual endpoint name.
Messaging and queueing
Dynatrace detects queue and topic listeners in your applications and identifies them based on the listener class name. See supported messaging services.
Queue listener services
Queue listener services tell you which queues or topics you're listening to. These are lightweight services that don't have response times. They tell you how many messages a queue or topic has dequeued; they don't tell you anything about message processing—messaging services do that.
If Dynatrace automatically detects an event-based message listener, a queue listener service is always followed by a messaging service, which gives you insight into message details. However, if you're just monitoring a queue, and not looking into message details, the queue listener service can exist on its own.
Messaging services (consumer services) process messages from a queue or topic. A messaging service is always preceded by a queue listener service, which listens to the queue or topic the message came from.
If your application uses non-event-based message queue handlers or dequeues messages in batch style, Dynatrace can't automatically detect how the messages are processed. To get insight into that, you must create a custom service for the message processing.
Remoting services are divided into two categories:
- Remote method invocation (RMI)
- Remote procedure call (RPC)
For a list of remoting services supported by Dynatrace, see supported remoting services.
In the world of Java, Remote Method Invocation (RMI) is a common means of communication used by JVMs. As there may be many dynamically-created RMI servers within a single JVM, Dynatrace creates only a single RMI service for each process group. However, this doesn't mean that you lose visibility into your RMI services; Dynatrace tracks and monitors each RMI class as a separate request type.
Dynatrace tracks remote procedure calls by SDKs, Akka, and AWSLambda. Unlike RMI, Dynatrace creates a separate RPC service for each service endpoint.
Background activity services
In many cases services are called by threads that run in the background of your application or another application. These requests, executed in background threads, represent the background activity of monitored process groups that make calls to other services. They also track outgoing messages in queues.
For example, if you have a background thread in a Tomcat that makes web requests to Apache, Dynatrace represents this as an background activity service of Tomcat. You'll be able to see which requests Tomcat is making to your Apache by analyzing the response time of the background activity service of Tomcat.
Custom services allow you to instrument an application that is not built on standard technologies. You can also fine-tune your system, and instrument a particular method, class, or interface you're interested in. You can create custom services for Java, .NET, and PHP.
To learn how to create custom services, see Define custom services.
O* default service
The * default service is created when Dynatrace detects a service call to a span attachment. This is just a placeholder service that's required to build a valid tree structure. Note that Dynatrace currently doesn't support service detection for spans.
Adjusting service detection
Web server naming issues
In some cases, web servers don't have well-defined virtual hosts, server names, or sites. A web server may simply be named
localhost. This can result in multiple similar services that contain multiple web site instances. To remedy such issues, adjust process-group detection settings.
When there is no virtual host configured in an Apache HTTP server, the web server name defaults to the name of the physical host. In cloud environments, this leads to one virtual host for each physical host instance and thus one service instance. If the cloud environment starts and stops the hosts, these services will be temporary.
To remedy such localhost scenarios, use an environment variable to define virtual host names. Simply set
DT_LOCALTOVIRTUALHOSTNAME for each web server process to any value. Dynatrace will pick up the names and use them in place of the existing localhost virtual host names. With this approach, you can ensure that multiple physical hosts report the same virtual host and thus get one service with multiple instances, one instance per physical host.
Define web application IDs
Some technologies don't provide unique application names. In such cases, you can define an environment variable called
DT_APPLICATIONID to provide a unique name. This only impacts services of the respective process that don't already have application IDs. For Java applications, you can alternatively use the system property
Rotating and anonymous ports
Dynatrace takes the listen port of each web request service into account when naming and detecting requests. In some cases, these ports are meaningless or random, changing with each restart. This is especially true if you're using a load balancer that dynamically assigns ports to application processes, as is the case in many Node.js scenarios.
To remedy this, you can set the environment variable
DT_IGNOREDYNAMICPORT=true. This removes the port from the detection and replaces it with